diff --git a/.github/workflows/build_and_deploy.yml b/.github/workflows/build_and_deploy.yml
index 28f246220..95eefcc69 100644
--- a/.github/workflows/build_and_deploy.yml
+++ b/.github/workflows/build_and_deploy.yml
@@ -228,7 +228,7 @@ jobs:
       - name: Raplace resource
         if: inputs.is_production
         shell: bash
-        run: |
+        run:
           mv -f download/resource/core/README.md ./README.md
       - name: Install cargo-binstall
         uses: taiki-e/install-action@cargo-binstall
@@ -298,7 +298,7 @@ jobs:
           cp -v target/${{ matrix.target }}/release/libvoicevox_core_java_api.so java_artifact/ || true
       - name: Code signing (Windows)
         if: startsWith(matrix.os, 'windows') && inputs.code_signing
-        run: |
+        run:
           bash build_util/codesign.bash "artifact/${{ env.ASSET_NAME }}/voicevox_core.dll"
         env:
           ESIGNERCKA_USERNAME: ${{ secrets.ESIGNERCKA_USERNAME }}
@@ -344,44 +344,34 @@ jobs:
     needs: [config, build_and_deploy]
     runs-on: macos-12
     env:
+      IOS_X86_64_PATH: artifact/voicevox_core-x86_64-apple-ios
+      IOS_AARCH64_SIM_PATH: artifact/voicevox_core-aarch64-apple-ios-sim
+      IOS_AARCH64_PATH: artifact/voicevox_core-aarch64-apple-ios
       ASSET_NAME: voicevox_core-ios-xcframework-cpu-${{ needs.config.outputs.version }}
     steps:
       - uses: actions/checkout@v3
       - uses: actions/download-artifact@v2
         with:
           name: voicevox_core-x86_64-apple-ios
-          path: artifact/voicevox_core-x86_64-apple-ios
+          path: ${{ env.IOS_X86_64_PATH }}
       - uses: actions/download-artifact@v2
         with:
           name: voicevox_core-aarch64-apple-ios-sim
-          path: artifact/voicevox_core-aarch64-apple-ios-sim
+          path: ${{ env.IOS_AARCH64_SIM_PATH }}
       - uses: actions/download-artifact@v2
         with:
           name: voicevox_core-aarch64-apple-ios
-          path: artifact/voicevox_core-aarch64-apple-ios
-      - name: Create fat binary
+          path: ${{ env.IOS_AARCH64_PATH }}
+      - name: Create xcframework
+        id: create-xcframework 
         run: |
-          mkdir -p "artifact/voicevox_core-sim"
-          lipo -create "artifact/voicevox_core-x86_64-apple-ios/libvoicevox_core.dylib" "artifact/voicevox_core-aarch64-apple-ios-sim/libvoicevox_core.dylib" -output "artifact/voicevox_core-sim/libvoicevox_core.dylib"
-      - name: Create XCFramework
-        run: |
-          mkdir -p "artifact/${{ env.ASSET_NAME }}"
-          # 必要なファイルだけコピー
-          mkdir -p "Headers-sim"
-          cp -v artifact/voicevox_core-x86_64-apple-ios/voicevox_core.h "Headers-sim"
-          cp -v crates/voicevox_core_c_api/xcframework/Headers/module.modulemap "Headers-sim"
-          mkdir -p "Headers-aarch64"
-          cp -v artifact/voicevox_core-aarch64-apple-ios/voicevox_core.h "Headers-aarch64"
-          cp -v crates/voicevox_core_c_api/xcframework/Headers/module.modulemap "Headers-aarch64"
-          xcodebuild -create-xcframework \
-            -library "artifact/voicevox_core-sim/libvoicevox_core.dylib" \
-            -headers "Headers-sim" \
-            -library "artifact/voicevox_core-aarch64-apple-ios/libvoicevox_core.dylib" \
-            -headers "Headers-aarch64" \
-            -output "artifact/${{ env.ASSET_NAME }}/voicevox_core.xcframework"
+          build_util/make_ios_xcframework.bash
+          echo "output_asset_path=${OUTPUT_ASSET_PATH}" >> "$GITHUB_OUTPUT" 
+        env:
+          OUTPUT_ASSET_PATH: artifact/voicevox_core-ios-xcframework-cpu
       - name: Archive artifact
         run: |
-          cd artifact/${{ env.ASSET_NAME }}
+          cd ${{ steps.create-xcframework.outputs.output_asset_path }}
           7z a "../../${{ env.ASSET_NAME }}.zip" "voicevox_core.xcframework"
       - name: Upload to Release
         if: fromJson(needs.config.outputs.deploy)
@@ -410,7 +400,7 @@ jobs:
       - name: Raplace resource
         if: inputs.is_production
         shell: bash
-        run: |
+        run:
           rm -r ./model; mv download/fat_resource/core/model ./model
       - name: Create artifact
         run: |
@@ -452,7 +442,7 @@ jobs:
       - name: Install cargo-edit
         run: cargo binstall cargo-edit@^0.11 --no-confirm
       - name: set cargo version
-        run: |
+        run:
           cargo set-version "$VERSION" -p voicevox_core_java_api
 
       - name: "Download artifact (android-arm64-cpu)"
diff --git a/.github/workflows/cargo-deny.yml b/.github/workflows/cargo-deny.yml
index ac4978b9a..cf600f535 100644
--- a/.github/workflows/cargo-deny.yml
+++ b/.github/workflows/cargo-deny.yml
@@ -13,7 +13,7 @@ jobs:
       - name: Install cargo-binstall
         uses: taiki-e/install-action@cargo-binstall
       - name: Install cargo-deny
-        run: cargo binstall cargo-deny@^0.13 --no-confirm --log-level debug
+        run: cargo binstall cargo-deny@^0.14 --no-confirm --log-level debug
       - name: cargo-deny
         run: |
           if ${{ !!github.event.release }}; then
diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml
index aebeb0318..b409ea567 100644
--- a/.github/workflows/test.yml
+++ b/.github/workflows/test.yml
@@ -151,7 +151,8 @@ jobs:
           metadata=$(cargo metadata --format-version 1)
           version=$(
               jq -r '
-                (.workspace_members[] | select(startswith("xtask "))) as $xtask
+                .workspace_members as $workspace_members
+                  | (.packages[] | select(.name == "xtask").id | select(. as $id | $workspace_members | index($id))) as $xtask
                   | (.resolve.nodes[] | select(.id == $xtask).deps[] | select(.name == "cbindgen").pkg) as $cbindgen
                   | .packages[] | select(.id == $cbindgen).version
                 ' <<< "$metadata"
diff --git a/Cargo.lock b/Cargo.lock
index 824a492c3..9e2929033 100644
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -219,19 +219,15 @@ dependencies = [
 
 [[package]]
 name = "async-compression"
-version = "0.3.15"
+version = "0.4.6"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "942c7cd7ae39e91bde4820d74132e9862e62c2f386c3aa90ccf55949f5bad63a"
+checksum = "a116f46a969224200a0a97f29cfd4c50e7534e4b4826bd23ea2c3c533039c82c"
 dependencies = [
- "bzip2",
  "flate2",
  "futures-core",
+ "futures-io",
  "memchr",
  "pin-project-lite",
- "tokio",
- "xz2",
- "zstd",
- "zstd-safe",
 ]
 
 [[package]]
@@ -242,8 +238,8 @@ checksum = "871f9bb5e0a22eeb7e8cf16641feb87c9dc67032ccf8ff49e772eb9941d3a965"
 dependencies = [
  "async-task",
  "concurrent-queue",
- "fastrand",
- "futures-lite",
+ "fastrand 1.8.0",
+ "futures-lite 1.12.0",
  "once_cell",
  "slab",
 ]
@@ -259,7 +255,7 @@ dependencies = [
  "async-io",
  "async-lock",
  "blocking",
- "futures-lite",
+ "futures-lite 1.12.0",
  "once_cell",
 ]
 
@@ -271,7 +267,7 @@ checksum = "83e21f3a490c72b3b0cf44962180e60045de2925d8dff97918f7ee43c8f637c7"
 dependencies = [
  "autocfg",
  "concurrent-queue",
- "futures-lite",
+ "futures-lite 1.12.0",
  "libc",
  "log",
  "once_cell",
@@ -307,7 +303,7 @@ dependencies = [
  "futures-channel",
  "futures-core",
  "futures-io",
- "futures-lite",
+ "futures-lite 1.12.0",
  "gloo-timers",
  "kv-log-macro",
  "log",
@@ -338,17 +334,15 @@ dependencies = [
 
 [[package]]
 name = "async_zip"
-version = "0.0.11"
+version = "0.0.16"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c50d29ab7e2f9e808cca1a69ea56a36f4ff216f54a41a23aae1fd4afc05cc020"
+checksum = "527207465fb6dcafbf661b0d4a51d0d2306c9d0c2975423079a6caa807930daf"
 dependencies = [
  "async-compression",
- "chrono",
  "crc32fast",
- "log",
+ "futures-lite 2.2.0",
  "pin-project",
  "thiserror",
- "tokio",
 ]
 
 [[package]]
@@ -479,8 +473,8 @@ dependencies = [
  "async-channel",
  "async-task",
  "atomic-waker",
- "fastrand",
- "futures-lite",
+ "fastrand 1.8.0",
+ "futures-lite 1.12.0",
  "once_cell",
 ]
 
@@ -1094,7 +1088,7 @@ dependencies = [
  "proc-macro2",
  "quote",
  "strsim",
- "syn 2.0.38",
+ "syn 2.0.48",
 ]
 
 [[package]]
@@ -1105,7 +1099,7 @@ checksum = "836a9bbc7ad63342d6d6e7b815ccab164bc77a2d95d84bc3117a8c0d5c98e2d5"
 dependencies = [
  "darling_core",
  "quote",
- "syn 2.0.38",
+ "syn 2.0.48",
 ]
 
 [[package]]
@@ -1292,7 +1286,7 @@ checksum = "48c69b3965971f5d0ea6a6dd26b55cdd517ae0e1425dc8d94e482a5915bd7ddf"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.38",
+ "syn 2.0.48",
 ]
 
 [[package]]
@@ -1305,7 +1299,7 @@ dependencies = [
  "num-traits",
  "proc-macro2",
  "quote",
- "syn 2.0.38",
+ "syn 2.0.48",
 ]
 
 [[package]]
@@ -1390,6 +1384,12 @@ dependencies = [
  "instant",
 ]
 
+[[package]]
+name = "fastrand"
+version = "2.0.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "25cbce373ec4653f1a01a31e8a5e5ec0c622dc27ff9c4e6606eefef5cbbed4a5"
+
 [[package]]
 name = "filetime"
 version = "0.2.17"
@@ -1502,7 +1502,7 @@ version = "1.12.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "7694489acd39452c77daa48516b894c153f192c3578d5a839b62c58099fcbf48"
 dependencies = [
- "fastrand",
+ "fastrand 1.8.0",
  "futures-core",
  "futures-io",
  "memchr",
@@ -1511,6 +1511,19 @@ dependencies = [
  "waker-fn",
 ]
 
+[[package]]
+name = "futures-lite"
+version = "2.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "445ba825b27408685aaecefd65178908c36c6e96aaf6d8599419d46e624192ba"
+dependencies = [
+ "fastrand 2.0.1",
+ "futures-core",
+ "futures-io",
+ "parking",
+ "pin-project-lite",
+]
+
 [[package]]
 name = "futures-macro"
 version = "0.3.28"
@@ -1519,7 +1532,7 @@ checksum = "89ca545a94061b6365f2c7355b4b32bd20df3ff95f02da9329b34ccc3bd6ee72"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.38",
+ "syn 2.0.48",
 ]
 
 [[package]]
@@ -1788,7 +1801,7 @@ dependencies = [
  "async-std",
  "base64 0.13.0",
  "cookie",
- "futures-lite",
+ "futures-lite 1.12.0",
  "infer",
  "pin-project-lite",
  "rand 0.7.3",
@@ -1934,15 +1947,15 @@ checksum = "cef509aa9bc73864d6756f0d34d35504af3cf0844373afe9b8669a5b8005a729"
 dependencies = [
  "console",
  "number_prefix",
- "portable-atomic",
+ "portable-atomic 0.3.19",
  "unicode-width",
 ]
 
 [[package]]
 name = "indoc"
-version = "1.0.7"
+version = "2.0.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "adab1eaa3408fb7f0c777a73e7465fd5656136fc93b670eb6df3c88c2c1344e3"
+checksum = "1e186cfbae8084e513daff4240b4797e342f988cecda4fb6c939150f96315fd8"
 
 [[package]]
 name = "infer"
@@ -2009,7 +2022,7 @@ dependencies = [
  "curl",
  "curl-sys",
  "flume",
- "futures-lite",
+ "futures-lite 1.12.0",
  "http",
  "log",
  "once_cell",
@@ -2045,6 +2058,15 @@ version = "1.0.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "4217ad341ebadf8d8e724e264f13e593e0648f5b3e94b3896a5df283be015ecc"
 
+[[package]]
+name = "jlabel"
+version = "0.1.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1f040b22c55628977296069dbf8635be49cc510999c048a1f1bdb56d00983148"
+dependencies = [
+ "thiserror",
+]
+
 [[package]]
 name = "jni"
 version = "0.21.1"
@@ -2216,17 +2238,6 @@ dependencies = [
  "value-bag",
 ]
 
-[[package]]
-name = "lzma-sys"
-version = "0.1.20"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "5fda04ab3764e6cde78b9974eec4f779acaba7c4e84b36eca3cf77c581b85d27"
-dependencies = [
- "cc",
- "libc",
- "pkg-config",
-]
-
 [[package]]
 name = "matchers"
 version = "0.1.0"
@@ -2591,7 +2602,7 @@ dependencies = [
  "proc-macro-error",
  "proc-macro2",
  "quote",
- "syn 2.0.38",
+ "syn 2.0.48",
 ]
 
 [[package]]
@@ -2617,9 +2628,9 @@ checksum = "c1b04fb49957986fdce4d6ee7a65027d55d4b6d2265e5848bbb507b58ccfdb6f"
 
 [[package]]
 name = "parking"
-version = "2.0.0"
+version = "2.2.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "427c3892f9e783d91cc128285287e70a59e206ca452770ece88a76f7a3eddd72"
+checksum = "bb813b8af86854136c6922af0598d719255ecb2179515e6e7730d468f05c9cae"
 
 [[package]]
 name = "parking_lot"
@@ -2667,7 +2678,7 @@ dependencies = [
  "regex",
  "regex-syntax 0.7.5",
  "structmeta",
- "syn 2.0.38",
+ "syn 2.0.48",
 ]
 
 [[package]]
@@ -2783,6 +2794,12 @@ version = "0.3.19"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "26f6a7b87c2e435a3241addceeeff740ff8b7e76b74c13bf9acb17fa454ea00b"
 
+[[package]]
+name = "portable-atomic"
+version = "1.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7170ef9988bc169ba16dd36a7fa041e5c4cbeb6a35b76d4c03daded371eae7c0"
+
 [[package]]
 name = "ppv-lite86"
 version = "0.2.16"
@@ -2862,9 +2879,9 @@ checksum = "dbf0c48bc1d91375ae5c3cd81e3722dff1abcf81a30960240640d223f59fe0e5"
 
 [[package]]
 name = "proc-macro2"
-version = "1.0.69"
+version = "1.0.78"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "134c189feb4956b20f6f547d2cf727d4c0fe06722b20a0eec87ed445a97f92da"
+checksum = "e2422ad645d89c99f8f3e6b88a9fdeca7fabeac836b1002371c4367c8f984aae"
 dependencies = [
  "unicode-ident",
 ]
@@ -2880,15 +2897,16 @@ dependencies = [
 
 [[package]]
 name = "pyo3"
-version = "0.19.2"
+version = "0.20.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e681a6cfdc4adcc93b4d3cf993749a4552018ee0a9b65fc0ccfad74352c72a38"
+checksum = "53bdbb96d49157e65d45cc287af5f32ffadd5f4761438b527b055fb0d4bb8233"
 dependencies = [
  "cfg-if",
  "indoc",
  "libc",
  "memoffset 0.9.0",
  "parking_lot",
+ "portable-atomic 1.6.0",
  "pyo3-build-config",
  "pyo3-ffi",
  "pyo3-macros",
@@ -2897,9 +2915,9 @@ dependencies = [
 
 [[package]]
 name = "pyo3-asyncio"
-version = "0.19.0"
+version = "0.20.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a2cc34c1f907ca090d7add03dc523acdd91f3a4dab12286604951e2f5152edad"
+checksum = "6ea6b68e93db3622f3bb3bf363246cf948ed5375afe7abff98ccbdd50b184995"
 dependencies = [
  "futures",
  "once_cell",
@@ -2910,9 +2928,9 @@ dependencies = [
 
 [[package]]
 name = "pyo3-build-config"
-version = "0.19.2"
+version = "0.20.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "076c73d0bc438f7a4ef6fdd0c3bb4732149136abd952b110ac93e4edb13a6ba5"
+checksum = "deaa5745de3f5231ce10517a1f5dd97d53e5a2fd77aa6b5842292085831d48d7"
 dependencies = [
  "once_cell",
  "target-lexicon",
@@ -2920,9 +2938,9 @@ dependencies = [
 
 [[package]]
 name = "pyo3-ffi"
-version = "0.19.2"
+version = "0.20.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e53cee42e77ebe256066ba8aa77eff722b3bb91f3419177cf4cd0f304d3284d9"
+checksum = "62b42531d03e08d4ef1f6e85a2ed422eb678b8cd62b762e53891c05faf0d4afa"
 dependencies = [
  "libc",
  "pyo3-build-config",
@@ -2941,32 +2959,34 @@ dependencies = [
 
 [[package]]
 name = "pyo3-macros"
-version = "0.19.2"
+version = "0.20.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "dfeb4c99597e136528c6dd7d5e3de5434d1ceaf487436a3f03b2d56b6fc9efd1"
+checksum = "7305c720fa01b8055ec95e484a6eca7a83c841267f0dd5280f0c8b8551d2c158"
 dependencies = [
  "proc-macro2",
  "pyo3-macros-backend",
  "quote",
- "syn 1.0.102",
+ "syn 2.0.48",
 ]
 
 [[package]]
 name = "pyo3-macros-backend"
-version = "0.19.2"
+version = "0.20.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "947dc12175c254889edc0c02e399476c2f652b4b9ebd123aa655c224de259536"
+checksum = "7c7e9b68bb9c3149c5b0cade5d07f953d6d125eb4337723c4ccdb665f1f96185"
 dependencies = [
+ "heck",
  "proc-macro2",
+ "pyo3-build-config",
  "quote",
- "syn 1.0.102",
+ "syn 2.0.48",
 ]
 
 [[package]]
 name = "quote"
-version = "1.0.33"
+version = "1.0.35"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "5267fca4496028628a95160fc423a33e8b2e6af8a5302579e322e4b520293cae"
+checksum = "291ec9ab5efd934aaf503a6466c5d5251535d108ee747472c3977cc5acc868ef"
 dependencies = [
  "proc-macro2",
 ]
@@ -3218,6 +3238,18 @@ dependencies = [
  "syn 1.0.102",
 ]
 
+[[package]]
+name = "rstest_reuse"
+version = "0.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "88530b681abe67924d42cca181d070e3ac20e0740569441a9e35a7cedd2b34a4"
+dependencies = [
+ "quote",
+ "rand 0.8.5",
+ "rustc_version 0.4.0",
+ "syn 2.0.48",
+]
+
 [[package]]
 name = "rustc-demangle"
 version = "0.1.21"
@@ -3397,7 +3429,7 @@ checksum = "d9735b638ccc51c28bf6914d90a2e9725b377144fc612c49a611fddd1b631d68"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.38",
+ "syn 2.0.48",
 ]
 
 [[package]]
@@ -3479,7 +3511,7 @@ dependencies = [
  "darling",
  "proc-macro2",
  "quote",
- "syn 2.0.38",
+ "syn 2.0.48",
 ]
 
 [[package]]
@@ -3591,9 +3623,9 @@ dependencies = [
 
 [[package]]
 name = "smallvec"
-version = "1.10.0"
+version = "1.13.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a507befe795404456341dfab10cef66ead4c041f62b8b11bbb92bffe5d0953e0"
+checksum = "e6ecd384b10a64542d77071bd64bd7b231f4ed5940fba55e98c3de13824cf3d7"
 
 [[package]]
 name = "snafu"
@@ -3722,7 +3754,7 @@ dependencies = [
  "proc-macro2",
  "quote",
  "structmeta-derive",
- "syn 2.0.38",
+ "syn 2.0.48",
 ]
 
 [[package]]
@@ -3733,7 +3765,7 @@ checksum = "a60bcaff7397072dca0017d1db428e30d5002e00b6847703e2e42005c95fbe00"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.38",
+ "syn 2.0.48",
 ]
 
 [[package]]
@@ -3800,9 +3832,9 @@ dependencies = [
 
 [[package]]
 name = "syn"
-version = "2.0.38"
+version = "2.0.48"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e96b79aaa137db8f61e26363a0c9b47d8b4ec75da28b7d1d614c2303e232408b"
+checksum = "0f3531638e407dfc0814761abb7c00a5b54992b849452a0646b7f65c9f770f3f"
 dependencies = [
  "proc-macro2",
  "quote",
@@ -3834,7 +3866,7 @@ checksum = "31c0432476357e58790aaa47a8efb0c5138f137343f3b5f23bd36a27e3b0a6d6"
 dependencies = [
  "autocfg",
  "cfg-if",
- "fastrand",
+ "fastrand 1.8.0",
  "redox_syscall 0.3.5",
  "rustix 0.37.19",
  "windows-sys 0.48.0",
@@ -3864,6 +3896,7 @@ dependencies = [
  "async_zip",
  "flate2",
  "fs-err",
+ "futures-lite 2.2.0",
  "once_cell",
  "serde",
  "serde_json",
@@ -3880,22 +3913,22 @@ checksum = "949517c0cf1bf4ee812e2e07e08ab448e3ae0d23472aee8a06c985f0c8815b16"
 
 [[package]]
 name = "thiserror"
-version = "1.0.37"
+version = "1.0.56"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "10deb33631e3c9018b9baf9dcbbc4f737320d2b576bac10f6aefa048fa407e3e"
+checksum = "d54378c645627613241d077a3a79db965db602882668f9136ac42af9ecb730ad"
 dependencies = [
  "thiserror-impl",
 ]
 
 [[package]]
 name = "thiserror-impl"
-version = "1.0.37"
+version = "1.0.56"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "982d17546b47146b28f7c22e3d08465f6b8903d0ea13c1660d9d84a6e7adcdbb"
+checksum = "fa0faa943b50f3db30a20aa7e265dbc66076993efed8463e8de414e5d06d3471"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 1.0.102",
+ "syn 2.0.48",
 ]
 
 [[package]]
@@ -4013,7 +4046,7 @@ checksum = "630bdcf245f78637c13ec01ffae6187cca34625e8c63150d424b59e55af2675e"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.38",
+ "syn 2.0.48",
 ]
 
 [[package]]
@@ -4247,9 +4280,9 @@ checksum = "c0edd1e5b14653f783770bce4a4dabb4a5108a5370a5f5d8cfe8710c361f6c8b"
 
 [[package]]
 name = "unindent"
-version = "0.1.10"
+version = "0.2.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "58ee9362deb4a96cef4d437d1ad49cffc9b9e92d202b6995674e928ce684f112"
+checksum = "c7de7d73e1754487cb58364ee906a499937a0dfabd86bcb980fa99ec8c8fa2ce"
 
 [[package]]
 name = "universal-hash"
@@ -4383,6 +4416,7 @@ dependencies = [
  "humansize",
  "indexmap 2.0.0",
  "itertools 0.10.5",
+ "jlabel",
  "nanoid",
  "ndarray",
  "once_cell",
@@ -4392,8 +4426,10 @@ dependencies = [
  "rayon",
  "regex",
  "rstest",
+ "rstest_reuse",
  "serde",
  "serde_json",
+ "smallvec",
  "tempfile",
  "test_util",
  "thiserror",
@@ -4471,7 +4507,7 @@ dependencies = [
  "indexmap 2.0.0",
  "proc-macro2",
  "quote",
- "syn 2.0.38",
+ "syn 2.0.48",
 ]
 
 [[package]]
@@ -4918,15 +4954,6 @@ dependencies = [
  "fs-err",
 ]
 
-[[package]]
-name = "xz2"
-version = "0.1.7"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "388c44dc09d76f1536602ead6d325eb532f5c122f17782bd57fb47baeeb767e2"
-dependencies = [
- "lzma-sys",
-]
-
 [[package]]
 name = "yansi"
 version = "0.5.1"
@@ -4980,11 +5007,10 @@ dependencies = [
 
 [[package]]
 name = "zstd-sys"
-version = "2.0.5+zstd.1.5.2"
+version = "2.0.9+zstd.1.5.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "edc50ffce891ad571e9f9afe5039c4837bede781ac4bb13052ed7ae695518596"
+checksum = "9e16efa8a874a0481a574084d34cc26fdb3b99627480f785888deb6386506656"
 dependencies = [
  "cc",
- "libc",
  "pkg-config",
 ]
diff --git a/Cargo.toml b/Cargo.toml
index e4e315514..2ed1092e5 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -9,7 +9,7 @@ anstyle-query = "1.0.0"
 anyhow = "1.0.65"
 assert_cmd = "2.0.8"
 async-std = "1.12.0"
-async_zip = "0.0.11"
+async_zip = "=0.0.16"
 binstall-tar = "0.4.39"
 bytes = "1.1.0"
 camino = "1.1.6"
@@ -33,12 +33,14 @@ fs-err = "2.9.0"
 futures = "0.3.26"
 futures-core = "0.3.25"
 futures-util = "0.3.25"
+futures-lite = "2.2.0"
 heck = "0.4.1"
 humansize = "2.1.2"
 indexmap = "2.0.0"
 indicatif = "0.17.3"
 inventory = "0.3.4"
 itertools = "0.10.5"
+jlabel = "0.1.2"
 jni = "0.21.1"
 libc = "0.2.134"
 libloading = "0.7.3"
@@ -53,17 +55,19 @@ ouroboros = "0.18.0"
 parse-display = "0.8.2"
 pretty_assertions = "1.3.0"
 proc-macro2 = "1.0.69"
-pyo3 = "0.19.2"
-pyo3-asyncio = "0.19.0"
+pyo3 = "0.20.3"
+pyo3-asyncio = "0.20.0"
 pyo3-log = "0.9.0"
 quote = "1.0.33"
 rayon = "1.6.1"
 regex = "1.10.0"
 reqwest = { version = "0.11.13", default-features = false }
 rstest = "0.15.0"
+rstest_reuse = "0.6.0"
 serde = "1.0.145"
 serde_json = "1.0.85"
 serde_with = "3.3.0"
+smallvec = "1.13.1"
 strum = "0.24.1"
 surf = "2.3.2"
 syn = "2.0.38"
diff --git a/build_util/make_ios_xcframework.bash b/build_util/make_ios_xcframework.bash
new file mode 100755
index 000000000..2d35fac88
--- /dev/null
+++ b/build_util/make_ios_xcframework.bash
@@ -0,0 +1,79 @@
+#!/usr/bin/env bash
+set -eu
+
+if [ ! -v IOS_X86_64_PATH ]; then # X86_64用のモジュールのディレクトリ(simulator)
+    echo "IOS_X86_64_PATHが未定義です"
+    exit 1
+fi
+if [ ! -v IOS_AARCH64_SIM_PATH ]; then # AARCH64_SIM用のモジュールのディレクトリ(simulator)
+    echo "IOS_AARCH64_SIM_PATHが未定義です"
+    exit 1
+fi
+if [ ! -v IOS_AARCH64_PATH ]; then # AARCH64用のモジュールのディレクトリ(実機)
+    echo "IOS_AARCH64_PATHが未定義です"
+    exit 1
+fi
+if [ ! -v OUTPUT_ASSET_PATH ]; then # 出力するASSETのディレクトリ
+    echo "OUTPUT_ASSET_PATHが未定義です"
+    exit 1
+fi
+
+echo "* Get original onnxruntime file name from rpath"
+output=$(otool -L "${IOS_AARCH64_PATH}/libvoicevox_core.dylib")
+matched_line=$(echo "$output" | grep "@rpath" | grep "libonnxruntime")
+if [[ $matched_line ]]; then
+    if [[ $matched_line =~ (@rpath/([^ ]+\.dylib)) ]]; then
+        dylib_string=${BASH_REMATCH[2]}
+    else
+        echo "Expected pattern not found in the matched line"
+        echo "$output"
+        exit 1
+    fi
+else
+    echo "No line containing '@rpath' and 'libonnxruntime' found"
+    echo "$output"
+    exit 1
+fi
+echo "Original onnx dylib file name: $dylib_string"
+
+echo "* Copy Framework template"
+arches=("aarch64" "sim")
+artifacts=("${IOS_AARCH64_PATH}" "${IOS_AARCH64_SIM_PATH}")
+for i in "${!arches[@]}"; do
+    arch="${arches[$i]}"
+    artifact="${artifacts[$i]}"
+    echo "* Copy Framework-${arch} template"
+    mkdir -p "Framework-${arch}/voicevox_core.framework/Headers"
+    cp -vr "crates/voicevox_core_c_api/xcframework/Frameworks/${arch}/" "Framework-${arch}/"
+    cp -v "${artifact}/voicevox_core.h" \
+        "Framework-${arch}/voicevox_core.framework/Headers/voicevox_core.h"
+done
+
+echo "* Create dylib"
+# aarch64はdylibをコピー
+cp -v "${IOS_AARCH64_PATH}/libvoicevox_core.dylib" \
+    "Framework-aarch64/voicevox_core.framework/voicevox_core"
+
+# simはx86_64とarrch64を合わせてdylib作成
+lipo -create "${IOS_X86_64_PATH}/libvoicevox_core.dylib" \
+    "${IOS_AARCH64_SIM_PATH}/libvoicevox_core.dylib" \
+    -output "Framework-sim/voicevox_core.framework/voicevox_core"
+
+for arch in "${arches[@]}"; do
+    echo "* Change ${arch} @rpath"
+    # 自身への@rpathを変更
+    install_name_tool -id "@rpath/voicevox_core.framework/voicevox_core" \
+        "Framework-${arch}/voicevox_core.framework/voicevox_core"
+
+    # 依存ライブラリonnxruntimeへの@rpathを変更
+    install_name_tool -change "@rpath/$dylib_string" \
+        "@rpath/onnxruntime.framework/onnxruntime" \
+        "Framework-${arch}/voicevox_core.framework/voicevox_core"
+done
+
+echo "* Create XCFramework"
+mkdir -p "${OUTPUT_ASSET_PATH}"
+xcodebuild -create-xcframework \
+    -framework "Framework-sim/voicevox_core.framework" \
+    -framework "Framework-aarch64/voicevox_core.framework" \
+    -output "${OUTPUT_ASSET_PATH}/voicevox_core.xcframework"
diff --git a/crates/test_util/Cargo.toml b/crates/test_util/Cargo.toml
index 6394b321a..bea6e4d4e 100644
--- a/crates/test_util/Cargo.toml
+++ b/crates/test_util/Cargo.toml
@@ -3,7 +3,8 @@ name = "test_util"
 edition.workspace = true
 
 [dependencies]
-async_zip = { workspace = true, features = ["full"] }
+async_zip = { workspace = true, features = ["deflate"] }
+futures-lite.workspace = true
 once_cell.workspace = true
 serde = { workspace = true, features = ["derive"] }
 serde_json.workspace = true
diff --git a/crates/test_util/src/lib.rs b/crates/test_util/src/lib.rs
index 05da2a33f..03e8309b8 100644
--- a/crates/test_util/src/lib.rs
+++ b/crates/test_util/src/lib.rs
@@ -1,6 +1,7 @@
 mod typing;
 
-use async_zip::{write::ZipFileWriter, Compression, ZipEntryBuilder};
+use async_zip::{base::write::ZipFileWriter, Compression, ZipEntryBuilder};
+use futures_lite::AsyncWriteExt as _;
 use once_cell::sync::Lazy;
 use std::{
     collections::HashMap,
@@ -8,7 +9,7 @@ use std::{
 };
 use tokio::{
     fs::{self, File},
-    io::{AsyncReadExt, AsyncWriteExt},
+    io::AsyncReadExt,
     sync::Mutex,
 };
 
@@ -21,7 +22,7 @@ pub const OPEN_JTALK_DIC_DIR: &str = concat!(
     "/data/open_jtalk_dic_utf_8-1.11"
 );
 
-pub const EXAMPLE_DATA_JSON: &str = include_str!(concat!(
+const EXAMPLE_DATA_JSON: &str = include_str!(concat!(
     env!("CARGO_MANIFEST_DIR"),
     "/data/example_data.json"
 ));
@@ -50,18 +51,11 @@ pub async fn convert_zip_vvm(dir: impl AsRef<Path>) -> PathBuf {
         fs::create_dir_all(out_file_path.parent().unwrap())
             .await
             .unwrap();
-        let mut out_file = File::create(&out_file_path).await.unwrap();
-        let mut writer = ZipFileWriter::new(&mut out_file);
+        let mut writer = ZipFileWriter::new(vec![]);
 
         for entry in dir.read_dir().unwrap().flatten() {
             let entry_builder = ZipEntryBuilder::new(
-                entry
-                    .path()
-                    .file_name()
-                    .unwrap()
-                    .to_str()
-                    .unwrap()
-                    .to_string(),
+                entry.path().file_name().unwrap().to_str().unwrap().into(),
                 Compression::Deflate,
             );
             let mut entry_writer = writer.write_entry_stream(entry_builder).await.unwrap();
@@ -71,7 +65,8 @@ pub async fn convert_zip_vvm(dir: impl AsRef<Path>) -> PathBuf {
             entry_writer.write_all(&buf).await.unwrap();
             entry_writer.close().await.unwrap();
         }
-        writer.close().await.unwrap();
+        let zip = writer.close().await.unwrap();
+        fs::write(&out_file_path, zip).await.unwrap();
     }
     out_file_path
 }
diff --git a/crates/voicevox_core/Cargo.toml b/crates/voicevox_core/Cargo.toml
index fab7e4956..ebe346823 100644
--- a/crates/voicevox_core/Cargo.toml
+++ b/crates/voicevox_core/Cargo.toml
@@ -11,7 +11,7 @@ directml = ["voicevox-ort/directml"]
 
 [dependencies]
 anyhow.workspace = true
-async_zip = { workspace = true, features = ["full"] }
+async_zip = { workspace = true, features = ["deflate"] }
 camino.workspace = true
 derive-getters.workspace = true
 derive-new.workspace = true
@@ -24,6 +24,7 @@ fs-err = { workspace = true, features = ["tokio"] }
 futures.workspace = true
 indexmap = { workspace = true, features = ["serde"] }
 itertools.workspace = true
+jlabel.workspace = true
 nanoid.workspace = true
 ndarray.workspace = true
 once_cell.workspace = true
@@ -33,6 +34,7 @@ rayon.workspace = true
 regex.workspace = true
 serde = { workspace = true, features = ["derive"] }
 serde_json = { workspace = true, features = ["preserve_order"] }
+smallvec.workspace = true
 tempfile.workspace = true
 thiserror.workspace = true
 tokio = { workspace = true, features = ["rt"] } # FIXME: feature-gateする
@@ -46,6 +48,7 @@ zip.workspace = true
 heck.workspace = true
 pretty_assertions.workspace = true
 rstest.workspace = true
+rstest_reuse.workspace = true
 test_util.workspace = true
 tokio = { workspace = true, features = ["rt", "macros"] }
 
diff --git a/crates/voicevox_core/src/engine/acoustic_feature_extractor.rs b/crates/voicevox_core/src/engine/acoustic_feature_extractor.rs
index 4a30f71a5..5ee7ea540 100644
--- a/crates/voicevox_core/src/engine/acoustic_feature_extractor.rs
+++ b/crates/voicevox_core/src/engine/acoustic_feature_extractor.rs
@@ -61,7 +61,7 @@ static PHONEME_MAP: Lazy<HashMap<&str, i64>> = Lazy::new(|| {
 });
 
 #[derive(Debug, Clone, PartialEq, new, Default, Getters)]
-pub struct OjtPhoneme {
+pub(crate) struct OjtPhoneme {
     phoneme: String,
     #[allow(dead_code)]
     start: f32,
@@ -70,15 +70,15 @@ pub struct OjtPhoneme {
 }
 
 impl OjtPhoneme {
-    pub fn num_phoneme() -> usize {
+    pub(crate) fn num_phoneme() -> usize {
         PHONEME_MAP.len()
     }
 
-    pub fn space_phoneme() -> String {
+    fn space_phoneme() -> String {
         "pau".into()
     }
 
-    pub fn phoneme_id(&self) -> i64 {
+    pub(crate) fn phoneme_id(&self) -> i64 {
         if self.phoneme.is_empty() {
             -1
         } else {
@@ -86,7 +86,7 @@ impl OjtPhoneme {
         }
     }
 
-    pub fn convert(phonemes: &[OjtPhoneme]) -> Vec<OjtPhoneme> {
+    pub(crate) fn convert(phonemes: &[OjtPhoneme]) -> Vec<OjtPhoneme> {
         let mut phonemes = phonemes.to_owned();
         if let Some(first_phoneme) = phonemes.first_mut() {
             if first_phoneme.phoneme.contains("sil") {
diff --git a/crates/voicevox_core/src/engine/full_context_label.rs b/crates/voicevox_core/src/engine/full_context_label.rs
index 79928d246..06cfdd0fd 100644
--- a/crates/voicevox_core/src/engine/full_context_label.rs
+++ b/crates/voicevox_core/src/engine/full_context_label.rs
@@ -1,10 +1,11 @@
-use std::collections::HashMap;
+use std::str::FromStr;
 
-use crate::engine::open_jtalk::FullcontextExtractor;
-use derive_getters::Getters;
-use derive_new::new;
-use once_cell::sync::Lazy;
-use regex::Regex;
+use crate::{
+    engine::{self, open_jtalk::FullcontextExtractor, MoraModel},
+    AccentPhraseModel,
+};
+use jlabel::{Label, Mora};
+use smallvec::SmallVec;
 
 // FIXME: 入力テキストをここで持って、メッセージに含む
 #[derive(thiserror::Error, Debug)]
@@ -20,321 +21,544 @@ enum ErrorKind {
     #[display(fmt = "Open JTalkで解釈することができませんでした")]
     OpenJtalk,
 
-    #[display(fmt = "label parse error label: {label}")]
-    LabelParse { label: String },
+    #[display(fmt = "jlabelでラベルを解釈することができませんでした")]
+    Jlabel,
 
-    #[display(fmt = "too long mora mora_phonemes: {mora_phonemes:?}")]
-    TooLongMora { mora_phonemes: Vec<Phoneme> },
-
-    #[display(fmt = "invalid mora: {mora:?}")]
-    InvalidMora { mora: Box<Mora> },
+    #[display(fmt = "too long mora")]
+    TooLongMora,
 }
 
 type Result<T> = std::result::Result<T, FullContextLabelError>;
 
-#[derive(new, Getters, Clone, PartialEq, Eq, Debug)]
-pub struct Phoneme {
-    contexts: HashMap<String, String>,
-    label: String,
-}
-
-static P3_REGEX: Lazy<Regex> = Lazy::new(|| Regex::new(r"(\-(.*?)\+)").unwrap());
-static A2_REGEX: Lazy<Regex> = Lazy::new(|| Regex::new(r"(\+(\d+|xx)\+)").unwrap());
-static A3_REGEX: Lazy<Regex> = Lazy::new(|| Regex::new(r"(\+(\d+|xx)/B:)").unwrap());
-static F1_REGEX: Lazy<Regex> = Lazy::new(|| Regex::new(r"(/F:(\d+|xx)_)").unwrap());
-static F2_REGEX: Lazy<Regex> = Lazy::new(|| Regex::new(r"(_(\d+|xx)\#)").unwrap());
-static F3_REGEX: Lazy<Regex> = Lazy::new(|| Regex::new(r"(\#(\d+|xx)_)").unwrap());
-static F5_REGEX: Lazy<Regex> = Lazy::new(|| Regex::new(r"(@(\d+|xx)_)").unwrap());
-static H1_REGEX: Lazy<Regex> = Lazy::new(|| Regex::new(r"(/H:(\d+|xx)_)").unwrap());
-static I3_REGEX: Lazy<Regex> = Lazy::new(|| Regex::new(r"(@(\d+|xx)\+)").unwrap());
-static J1_REGEX: Lazy<Regex> = Lazy::new(|| Regex::new(r"(/J:(\d+|xx)_)").unwrap());
-
-fn string_feature_by_regex(re: &Regex, label: &str) -> std::result::Result<String, ErrorKind> {
-    if let Some(caps) = re.captures(label) {
-        Ok(caps[2].to_string())
-    } else {
-        Err(ErrorKind::LabelParse {
-            label: label.into(),
-        })
-    }
-}
-
-impl Phoneme {
-    fn from_label(label: impl Into<String>) -> std::result::Result<Self, ErrorKind> {
-        let mut contexts = HashMap::<String, String>::with_capacity(10);
-        let label = label.into();
-        contexts.insert("p3".into(), string_feature_by_regex(&P3_REGEX, &label)?);
-        contexts.insert("a2".into(), string_feature_by_regex(&A2_REGEX, &label)?);
-        contexts.insert("a3".into(), string_feature_by_regex(&A3_REGEX, &label)?);
-        contexts.insert("f1".into(), string_feature_by_regex(&F1_REGEX, &label)?);
-        contexts.insert("f2".into(), string_feature_by_regex(&F2_REGEX, &label)?);
-        contexts.insert("f3".into(), string_feature_by_regex(&F3_REGEX, &label)?);
-        contexts.insert("f5".into(), string_feature_by_regex(&F5_REGEX, &label)?);
-        contexts.insert("h1".into(), string_feature_by_regex(&H1_REGEX, &label)?);
-        contexts.insert("i3".into(), string_feature_by_regex(&I3_REGEX, &label)?);
-        contexts.insert("j1".into(), string_feature_by_regex(&J1_REGEX, &label)?);
-
-        Ok(Self::new(contexts, label))
-    }
-
-    pub fn phoneme(&self) -> &str {
-        self.contexts.get("p3").unwrap().as_str()
-    }
-
-    pub fn is_pause(&self) -> bool {
-        self.contexts.get("f1").unwrap().as_str() == "xx"
-    }
-}
-
-#[derive(new, Getters, Clone, PartialEq, Eq, Debug)]
-pub struct Mora {
-    consonant: Option<Phoneme>,
-    vowel: Phoneme,
+pub(crate) fn extract_full_context_label(
+    open_jtalk: &impl FullcontextExtractor,
+    text: impl AsRef<str>,
+) -> Result<Vec<AccentPhraseModel>> {
+    let labels = open_jtalk
+        .extract_fullcontext(text.as_ref())
+        .map_err(|source| FullContextLabelError {
+            context: ErrorKind::OpenJtalk,
+            source: Some(source),
+        })?;
+
+    let parsed_labels = labels
+        .into_iter()
+        .map(|s| Label::from_str(&s))
+        .collect::<std::result::Result<Vec<_>, _>>()
+        .map_err(|source| FullContextLabelError {
+            context: ErrorKind::Jlabel,
+            source: Some(source.into()),
+        })?;
+
+    generate_accent_phrases(&parsed_labels).map_err(|context| FullContextLabelError {
+        context,
+        source: None,
+    })
 }
 
-impl Mora {
-    pub fn set_context(&mut self, key: impl Into<String>, value: impl Into<String>) {
-        let key = key.into();
-        let value = value.into();
-        if let Some(ref mut consonant) = self.consonant {
-            consonant.contexts.insert(key.clone(), value.clone());
+fn generate_accent_phrases(
+    utterance: &[Label],
+) -> std::result::Result<Vec<AccentPhraseModel>, ErrorKind> {
+    let mut accent_phrases = Vec::with_capacity(
+        utterance
+            .first()
+            .map(|label| label.utterance.accent_phrase_count.into())
+            .unwrap_or(0),
+    );
+
+    let split = utterance.chunk_by(|a, b| {
+        a.breath_group_curr == b.breath_group_curr && a.accent_phrase_curr == b.accent_phrase_curr
+    });
+    for labels in split {
+        let moras = generate_moras(labels)?;
+        if moras.is_empty() {
+            continue;
         }
-        self.vowel.contexts.insert(key, value);
-    }
 
-    pub fn phonemes(&self) -> Vec<Phoneme> {
-        if self.consonant.is_some() {
-            vec![
-                self.consonant().as_ref().unwrap().clone(),
-                self.vowel.clone(),
-            ]
+        let Some(Label {
+            accent_phrase_curr: Some(ap_curr),
+            breath_group_curr: Some(bg_curr),
+            ..
+        }) = labels.first()
+        else {
+            continue;
+        };
+
+        // Breath Groupの中で最後のアクセント句かつ，Utteranceの中で最後のBreath Groupでない場合は次がpauになる
+        let pause_mora = if ap_curr.accent_phrase_position_backward == 1
+            && bg_curr.breath_group_position_backward != 1
+        {
+            Some(MoraModel::new(
+                "、".into(),
+                None,
+                None,
+                "pau".into(),
+                0.,
+                0.,
+            ))
         } else {
-            vec![self.vowel.clone()]
-        }
-    }
+            None
+        };
 
-    #[allow(dead_code)]
-    pub fn labels(&self) -> Vec<String> {
-        self.phonemes().iter().map(|p| p.label().clone()).collect()
+        // workaround for VOICEVOX/voicevox_engine#55
+        let accent = usize::from(ap_curr.accent_position).min(moras.len());
+
+        accent_phrases.push(AccentPhraseModel::new(
+            moras,
+            accent,
+            pause_mora,
+            ap_curr.is_interrogative,
+        ))
     }
+    Ok(accent_phrases)
 }
 
-#[derive(new, Getters, Clone, Debug, PartialEq, Eq)]
-pub struct AccentPhrase {
-    moras: Vec<Mora>,
-    accent: usize,
-    is_interrogative: bool,
-}
-
-impl AccentPhrase {
-    fn from_phonemes(mut phonemes: Vec<Phoneme>) -> std::result::Result<Self, ErrorKind> {
-        let mut moras = Vec::with_capacity(phonemes.len());
-        let mut mora_phonemes = Vec::with_capacity(phonemes.len());
-        for i in 0..phonemes.len() {
-            {
-                let phoneme = phonemes.get_mut(i).unwrap();
-                if phoneme.contexts().get("a2").map(|s| s.as_str()) == Some("49") {
-                    break;
-                }
-                mora_phonemes.push(phoneme.clone());
+fn generate_moras(accent_phrase: &[Label]) -> std::result::Result<Vec<MoraModel>, ErrorKind> {
+    let mut moras = Vec::with_capacity(accent_phrase.len());
+
+    let split = accent_phrase.chunk_by(|a, b| a.mora == b.mora);
+    for labels in split {
+        let labels: SmallVec<[&Label; 3]> =
+            labels.iter().filter(|label| label.mora.is_some()).collect();
+        match labels[..] {
+            [consonant, vowel] => {
+                let mora = generate_mora(Some(consonant), vowel);
+                moras.push(mora);
             }
-
-            if i + 1 == phonemes.len()
-                || phonemes.get(i).unwrap().contexts().get("a2").unwrap()
-                    != phonemes.get(i + 1).unwrap().contexts().get("a2").unwrap()
-            {
-                if mora_phonemes.len() == 1 {
-                    moras.push(Mora::new(None, mora_phonemes[0].clone()));
-                } else if mora_phonemes.len() == 2 {
-                    moras.push(Mora::new(
-                        Some(mora_phonemes[0].clone()),
-                        mora_phonemes[1].clone(),
-                    ));
-                } else {
-                    return Err(ErrorKind::TooLongMora { mora_phonemes });
-                }
-                mora_phonemes.clear();
+            [vowel] => {
+                let mora = generate_mora(None, vowel);
+                moras.push(mora);
+            }
+            // silやpau以外の音素がないモーラは含めない
+            [] => {}
+
+            // 音素が3つ以上ある場合：
+            // position_forwardとposition_backwardが飽和している場合は無視する
+            [Label {
+                mora:
+                    Some(Mora {
+                        position_forward: 49,
+                        position_backward: 49,
+                        ..
+                    }),
+                ..
+            }, ..] => {}
+            _ => {
+                return Err(ErrorKind::TooLongMora);
             }
-        }
-
-        let mora = &moras[0];
-        let mut accent: usize = mora
-            .vowel()
-            .contexts()
-            .get("f2")
-            .ok_or_else(|| ErrorKind::InvalidMora {
-                mora: mora.clone().into(),
-            })?
-            .parse()
-            .map_err(|_| ErrorKind::InvalidMora {
-                mora: mora.clone().into(),
-            })?;
-
-        let is_interrogative = moras
-            .last()
-            .unwrap()
-            .vowel()
-            .contexts()
-            .get("f3")
-            .map(|s| s.as_str())
-            == Some("1");
-        // workaround for VOICEVOX/voicevox_engine#55
-        if accent > moras.len() {
-            accent = moras.len();
-        }
-
-        Ok(Self::new(moras, accent, is_interrogative))
-    }
-
-    #[allow(dead_code)]
-    pub fn set_context(&mut self, key: impl Into<String>, value: impl Into<String>) {
-        let key = key.into();
-        let value = value.into();
-        for mora in self.moras.iter_mut() {
-            mora.set_context(&key, &value);
         }
     }
-
-    pub fn phonemes(&self) -> Vec<Phoneme> {
-        self.moras.iter().flat_map(|m| m.phonemes()).collect()
-    }
-
-    #[allow(dead_code)]
-    pub fn labels(&self) -> Vec<String> {
-        self.phonemes().iter().map(|p| p.label().clone()).collect()
-    }
-
-    #[allow(dead_code)]
-    pub fn merge(&self, accent_phrase: AccentPhrase) -> AccentPhrase {
-        let mut moras = self.moras().clone();
-        let is_interrogative = *accent_phrase.is_interrogative();
-        moras.extend(accent_phrase.moras);
-        AccentPhrase::new(moras, *self.accent(), is_interrogative)
-    }
+    Ok(moras)
 }
 
-#[derive(new, Getters, Clone, PartialEq, Eq, Debug)]
-pub struct BreathGroup {
-    accent_phrases: Vec<AccentPhrase>,
+fn generate_mora(consonant: Option<&Label>, vowel: &Label) -> MoraModel {
+    let consonant_phoneme = consonant.and_then(|c| c.phoneme.c.to_owned());
+    let vowel_phoneme = vowel.phoneme.c.as_deref().unwrap();
+    MoraModel::new(
+        mora_to_text(consonant_phoneme.as_deref(), vowel_phoneme),
+        consonant_phoneme,
+        consonant.and(Some(0.0)),
+        vowel_phoneme.to_string(),
+        0.0,
+        0.0,
+    )
 }
 
-impl BreathGroup {
-    fn from_phonemes(phonemes: Vec<Phoneme>) -> std::result::Result<Self, ErrorKind> {
-        let mut accent_phrases = Vec::with_capacity(phonemes.len());
-        let mut accent_phonemes = Vec::with_capacity(phonemes.len());
-        for i in 0..phonemes.len() {
-            accent_phonemes.push(phonemes.get(i).unwrap().clone());
-            if i + 1 == phonemes.len()
-                || phonemes.get(i).unwrap().contexts().get("i3").unwrap()
-                    != phonemes.get(i + 1).unwrap().contexts().get("i3").unwrap()
-                || phonemes.get(i).unwrap().contexts().get("f5").unwrap()
-                    != phonemes.get(i + 1).unwrap().contexts().get("f5").unwrap()
-            {
-                accent_phrases.push(AccentPhrase::from_phonemes(accent_phonemes.clone())?);
-                accent_phonemes.clear();
-            }
+pub fn mora_to_text(consonant: Option<&str>, vowel: &str) -> String {
+    let mora_text = format!(
+        "{}{}",
+        consonant.unwrap_or(""),
+        match vowel {
+            phoneme @ ("A" | "I" | "U" | "E" | "O") => phoneme.to_lowercase(),
+            phoneme => phoneme.to_string(),
         }
+    );
+    // もしカタカナに変換できなければ、引数で与えた文字列がそのまま返ってくる
+    engine::mora2text(&mora_text).to_string()
+}
 
-        Ok(Self::new(accent_phrases))
+// FIXME: Rust 1.77の新機能導入と共にこれを消す
+#[allow(unused_imports)]
+mod chunk_by {
+    // Implementations in this module were copied from
+    // [Rust](https://github.com/rust-lang/rust/blob/746a58d4359786e4aebb372a30829706fa5a968f/library/core/src/slice/iter.rs).
+
+    // MIT License Notice
+
+    // Permission is hereby granted, free of charge, to any
+    // person obtaining a copy of this software and associated
+    // documentation files (the "Software"), to deal in the
+    // Software without restriction, including without
+    // limitation the rights to use, copy, modify, merge,
+    // publish, distribute, sublicense, and/or sell copies of
+    // the Software, and to permit persons to whom the Software
+    // is furnished to do so, subject to the following
+    // conditions:
+    //
+    // The above copyright notice and this permission notice
+    // shall be included in all copies or substantial portions
+    // of the Software.
+    //
+    // THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF
+    // ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED
+    // TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A
+    // PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT
+    // SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
+    // CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+    // OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR
+    // IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+    // DEALINGS IN THE SOFTWARE.
+
+    pub struct ChunkBy<'a, T, P> {
+        slice: &'a [T],
+        predicate: P,
     }
-
-    #[allow(dead_code)]
-    pub fn set_context(&mut self, key: impl Into<String>, value: impl Into<String>) {
-        let key = key.into();
-        let value = value.into();
-        for accent_phrase in self.accent_phrases.iter_mut() {
-            accent_phrase.set_context(&key, &value);
+    impl<'a, T, P> ChunkBy<'a, T, P> {
+        pub(super) fn new(slice: &'a [T], predicate: P) -> Self {
+            ChunkBy { slice, predicate }
         }
     }
+    impl<'a, T, P> Iterator for ChunkBy<'a, T, P>
+    where
+        P: FnMut(&T, &T) -> bool,
+    {
+        type Item = &'a [T];
+
+        #[inline]
+        fn next(&mut self) -> Option<Self::Item> {
+            if self.slice.is_empty() {
+                None
+            } else {
+                let mut len = 1;
+                let mut iter = self.slice.windows(2);
+                while let Some([l, r]) = iter.next() {
+                    if (self.predicate)(l, r) {
+                        len += 1
+                    } else {
+                        break;
+                    }
+                }
+                let (head, tail) = self.slice.split_at(len);
+                self.slice = tail;
+                Some(head)
+            }
+        }
 
-    pub fn phonemes(&self) -> Vec<Phoneme> {
-        self.accent_phrases()
-            .iter()
-            .flat_map(|a| a.phonemes())
-            .collect()
+        #[inline]
+        fn size_hint(&self) -> (usize, Option<usize>) {
+            if self.slice.is_empty() {
+                (0, Some(0))
+            } else {
+                (1, Some(self.slice.len()))
+            }
+        }
     }
 
-    #[allow(dead_code)]
-    pub fn labels(&self) -> Vec<String> {
-        self.phonemes().iter().map(|p| p.label().clone()).collect()
+    #[easy_ext::ext(TChunkBy)]
+    impl<T> [T] {
+        pub fn chunk_by<F>(&self, pred: F) -> ChunkBy<'_, T, F>
+        where
+            F: FnMut(&T, &T) -> bool,
+        {
+            ChunkBy::new(self, pred)
+        }
     }
-}
 
-#[derive(new, Getters, Clone, PartialEq, Eq, Debug)]
-pub struct Utterance {
-    breath_groups: Vec<BreathGroup>,
-    pauses: Vec<Phoneme>,
+    #[cfg(test)]
+    mod tests {
+        use super::TChunkBy;
+
+        #[test]
+        fn chunk_by() {
+            let mut split = [0, 0, 1, 1, 1, -5].chunk_by(|a, b| a == b);
+            assert_eq!(split.next(), Some([0, 0].as_slice()));
+            assert_eq!(split.next(), Some([1, 1, 1].as_slice()));
+            assert_eq!(split.next(), Some([-5].as_slice()));
+            assert_eq!(split.next(), None);
+        }
+    }
 }
 
-impl Utterance {
-    fn from_phonemes(phonemes: Vec<Phoneme>) -> std::result::Result<Self, ErrorKind> {
-        let mut breath_groups = vec![];
-        let mut group_phonemes = Vec::with_capacity(phonemes.len());
-        let mut pauses = vec![];
-        for phoneme in phonemes.into_iter() {
-            if !phoneme.is_pause() {
-                group_phonemes.push(phoneme);
-            } else {
-                pauses.push(phoneme);
-
-                if !group_phonemes.is_empty() {
-                    breath_groups.push(BreathGroup::from_phonemes(group_phonemes.clone())?);
-                    group_phonemes.clear();
-                }
-            }
-        }
-        Ok(Self::new(breath_groups, pauses))
+#[cfg(test)]
+mod tests {
+    use rstest_reuse::*;
+
+    use ::test_util::OPEN_JTALK_DIC_DIR;
+    use rstest::rstest;
+
+    use std::str::FromStr;
+
+    use crate::{
+        engine::{
+            full_context_label::{extract_full_context_label, generate_accent_phrases},
+            open_jtalk::FullcontextExtractor,
+            MoraModel,
+        },
+        AccentPhraseModel,
+    };
+    use jlabel::Label;
+
+    fn mora(text: &str, consonant: Option<&str>, vowel: &str) -> MoraModel {
+        MoraModel::new(
+            text.into(),
+            consonant.map(|c| c.into()),
+            consonant.and(Some(0.0)),
+            vowel.into(),
+            0.0,
+            0.0,
+        )
     }
 
-    #[allow(dead_code)]
-    pub fn set_context(&mut self, key: impl Into<String>, value: impl Into<String>) {
-        let key = key.into();
-        let value = value.into();
-        for breath_group in self.breath_groups.iter_mut() {
-            breath_group.set_context(&key, &value);
-        }
+    #[template]
+    #[rstest]
+    #[case(
+        "いぇ",
+        &[
+            "xx^xx-sil+y=e/A:xx+xx+xx/B:xx-xx_xx/C:xx_xx+xx/D:xx+xx_xx/E:xx_xx!xx_xx-xx/F:xx_xx#xx_xx@xx_xx|xx_xx/G:1_1%0_xx_xx/H:xx_xx/I:xx-xx@xx+xx&xx-xx|xx+xx/J:1_1/K:1+1-1",
+            "xx^sil-y+e=sil/A:0+1+1/B:xx-xx_xx/C:09_xx+xx/D:xx+xx_xx/E:xx_xx!xx_xx-xx/F:1_1#0_xx@1_1|1_1/G:xx_xx%xx_xx_xx/H:xx_xx/I:1-1@1+1&1-1|1+1/J:xx_xx/K:1+1-1",
+            "sil^y-e+sil=xx/A:0+1+1/B:xx-xx_xx/C:09_xx+xx/D:xx+xx_xx/E:xx_xx!xx_xx-xx/F:1_1#0_xx@1_1|1_1/G:xx_xx%xx_xx_xx/H:xx_xx/I:1-1@1+1&1-1|1+1/J:xx_xx/K:1+1-1",
+            "y^e-sil+xx=xx/A:xx+xx+xx/B:xx-xx_xx/C:xx_xx+xx/D:xx+xx_xx/E:1_1!0_xx-xx/F:xx_xx#xx_xx@xx_xx|xx_xx/G:xx_xx%xx_xx_xx/H:1_1/I:xx-xx@xx+xx&xx-xx|xx+xx/J:xx_xx/K:1+1-1",
+        ],
+        &[
+            AccentPhraseModel::new(
+                vec![mora("イェ", Some("y"), "e")],
+                1,
+                None,
+                false,
+            )
+        ]
+    )]
+    #[case(
+        "んーっ",
+        &[
+            "xx^xx-sil+N=N/A:xx+xx+xx/B:xx-xx_xx/C:xx_xx+xx/D:09+xx_xx/E:xx_xx!xx_xx-xx/F:xx_xx#xx_xx@xx_xx|xx_xx/G:3_3%0_xx_xx/H:xx_xx/I:xx-xx@xx+xx&xx-xx|xx+xx/J:1_3/K:1+1-3",
+            "xx^sil-N+N=cl/A:-2+1+3/B:xx-xx_xx/C:09_xx+xx/D:09+xx_xx/E:xx_xx!xx_xx-xx/F:3_3#0_xx@1_1|1_3/G:xx_xx%xx_xx_xx/H:xx_xx/I:1-3@1+1&1-1|1+3/J:xx_xx/K:1+1-3",
+            "sil^N-N+cl=sil/A:-1+2+2/B:xx-xx_xx/C:09_xx+xx/D:09+xx_xx/E:xx_xx!xx_xx-xx/F:3_3#0_xx@1_1|1_3/G:xx_xx%xx_xx_xx/H:xx_xx/I:1-3@1+1&1-1|1+3/J:xx_xx/K:1+1-3",
+            "N^N-cl+sil=xx/A:0+3+1/B:09-xx_xx/C:09_xx+xx/D:xx+xx_xx/E:xx_xx!xx_xx-xx/F:3_3#0_xx@1_1|1_3/G:xx_xx%xx_xx_xx/H:xx_xx/I:1-3@1+1&1-1|1+3/J:xx_xx/K:1+1-3",
+            "N^cl-sil+xx=xx/A:xx+xx+xx/B:09-xx_xx/C:xx_xx+xx/D:xx+xx_xx/E:3_3!0_xx-xx/F:xx_xx#xx_xx@xx_xx|xx_xx/G:xx_xx%xx_xx_xx/H:1_3/I:xx-xx@xx+xx&xx-xx|xx+xx/J:xx_xx/K:1+1-3",
+        ],
+        &[
+            AccentPhraseModel::new(
+                vec![
+                    mora("ン", None, "N"),
+                    mora("ン", None, "N"),
+                    mora("ッ", None, "cl"),
+                ],
+                3,
+                None,
+                false,
+            ),
+        ]
+    )]
+    #[case(
+        "これはテストです",
+        &[
+            "xx^xx-sil+k=o/A:xx+xx+xx/B:xx-xx_xx/C:xx_xx+xx/D:04+xx_xx/E:xx_xx!xx_xx-xx/F:xx_xx#xx_xx@xx_xx|xx_xx/G:3_3%0_xx_xx/H:xx_xx/I:xx-xx@xx+xx&xx-xx|xx+xx/J:2_8/K:1+2-8",
+            "xx^sil-k+o=r/A:-2+1+3/B:xx-xx_xx/C:04_xx+xx/D:24+xx_xx/E:xx_xx!xx_xx-xx/F:3_3#0_xx@1_2|1_8/G:5_1%0_xx_1/H:xx_xx/I:2-8@1+1&1-2|1+8/J:xx_xx/K:1+2-8",
+            "sil^k-o+r=e/A:-2+1+3/B:xx-xx_xx/C:04_xx+xx/D:24+xx_xx/E:xx_xx!xx_xx-xx/F:3_3#0_xx@1_2|1_8/G:5_1%0_xx_1/H:xx_xx/I:2-8@1+1&1-2|1+8/J:xx_xx/K:1+2-8",
+            "k^o-r+e=w/A:-1+2+2/B:xx-xx_xx/C:04_xx+xx/D:24+xx_xx/E:xx_xx!xx_xx-xx/F:3_3#0_xx@1_2|1_8/G:5_1%0_xx_1/H:xx_xx/I:2-8@1+1&1-2|1+8/J:xx_xx/K:1+2-8",
+            "o^r-e+w=a/A:-1+2+2/B:xx-xx_xx/C:04_xx+xx/D:24+xx_xx/E:xx_xx!xx_xx-xx/F:3_3#0_xx@1_2|1_8/G:5_1%0_xx_1/H:xx_xx/I:2-8@1+1&1-2|1+8/J:xx_xx/K:1+2-8",
+            "r^e-w+a=t/A:0+3+1/B:04-xx_xx/C:24_xx+xx/D:03+xx_xx/E:xx_xx!xx_xx-xx/F:3_3#0_xx@1_2|1_8/G:5_1%0_xx_1/H:xx_xx/I:2-8@1+1&1-2|1+8/J:xx_xx/K:1+2-8",
+            "e^w-a+t=e/A:0+3+1/B:04-xx_xx/C:24_xx+xx/D:03+xx_xx/E:xx_xx!xx_xx-xx/F:3_3#0_xx@1_2|1_8/G:5_1%0_xx_1/H:xx_xx/I:2-8@1+1&1-2|1+8/J:xx_xx/K:1+2-8",
+            "w^a-t+e=s/A:0+1+5/B:24-xx_xx/C:03_xx+xx/D:10+7_2/E:3_3!0_xx-1/F:5_1#0_xx@2_1|4_5/G:xx_xx%xx_xx_xx/H:xx_xx/I:2-8@1+1&1-2|1+8/J:xx_xx/K:1+2-8",
+            "a^t-e+s=U/A:0+1+5/B:24-xx_xx/C:03_xx+xx/D:10+7_2/E:3_3!0_xx-1/F:5_1#0_xx@2_1|4_5/G:xx_xx%xx_xx_xx/H:xx_xx/I:2-8@1+1&1-2|1+8/J:xx_xx/K:1+2-8",
+            "t^e-s+U=t/A:1+2+4/B:24-xx_xx/C:03_xx+xx/D:10+7_2/E:3_3!0_xx-1/F:5_1#0_xx@2_1|4_5/G:xx_xx%xx_xx_xx/H:xx_xx/I:2-8@1+1&1-2|1+8/J:xx_xx/K:1+2-8",
+            "e^s-U+t=o/A:1+2+4/B:24-xx_xx/C:03_xx+xx/D:10+7_2/E:3_3!0_xx-1/F:5_1#0_xx@2_1|4_5/G:xx_xx%xx_xx_xx/H:xx_xx/I:2-8@1+1&1-2|1+8/J:xx_xx/K:1+2-8",
+            "s^U-t+o=d/A:2+3+3/B:24-xx_xx/C:03_xx+xx/D:10+7_2/E:3_3!0_xx-1/F:5_1#0_xx@2_1|4_5/G:xx_xx%xx_xx_xx/H:xx_xx/I:2-8@1+1&1-2|1+8/J:xx_xx/K:1+2-8",
+            "U^t-o+d=e/A:2+3+3/B:24-xx_xx/C:03_xx+xx/D:10+7_2/E:3_3!0_xx-1/F:5_1#0_xx@2_1|4_5/G:xx_xx%xx_xx_xx/H:xx_xx/I:2-8@1+1&1-2|1+8/J:xx_xx/K:1+2-8",
+            "t^o-d+e=s/A:3+4+2/B:03-xx_xx/C:10_7+2/D:xx+xx_xx/E:3_3!0_xx-1/F:5_1#0_xx@2_1|4_5/G:xx_xx%xx_xx_xx/H:xx_xx/I:2-8@1+1&1-2|1+8/J:xx_xx/K:1+2-8",
+            "o^d-e+s=U/A:3+4+2/B:03-xx_xx/C:10_7+2/D:xx+xx_xx/E:3_3!0_xx-1/F:5_1#0_xx@2_1|4_5/G:xx_xx%xx_xx_xx/H:xx_xx/I:2-8@1+1&1-2|1+8/J:xx_xx/K:1+2-8",
+            "d^e-s+U=sil/A:4+5+1/B:03-xx_xx/C:10_7+2/D:xx+xx_xx/E:3_3!0_xx-1/F:5_1#0_xx@2_1|4_5/G:xx_xx%xx_xx_xx/H:xx_xx/I:2-8@1+1&1-2|1+8/J:xx_xx/K:1+2-8",
+            "e^s-U+sil=xx/A:4+5+1/B:03-xx_xx/C:10_7+2/D:xx+xx_xx/E:3_3!0_xx-1/F:5_1#0_xx@2_1|4_5/G:xx_xx%xx_xx_xx/H:xx_xx/I:2-8@1+1&1-2|1+8/J:xx_xx/K:1+2-8",
+            "s^U-sil+xx=xx/A:xx+xx+xx/B:10-7_2/C:xx_xx+xx/D:xx+xx_xx/E:5_1!0_xx-xx/F:xx_xx#xx_xx@xx_xx|xx_xx/G:xx_xx%xx_xx_xx/H:2_8/I:xx-xx@xx+xx&xx-xx|xx+xx/J:xx_xx/K:1+2-8",
+        ],
+        &[
+            AccentPhraseModel::new(
+                vec![
+                    mora("コ", Some("k"), "o"),
+                    mora("レ", Some("r"), "e"),
+                    mora("ワ", Some("w"), "a"),
+                ],
+                3,
+                None,
+                false,
+            ),
+            AccentPhraseModel::new(
+                vec![
+                    mora("テ", Some("t"), "e"),
+                    mora("ス", Some("s"), "U"),
+                    mora("ト", Some("t"), "o"),
+                    mora("デ", Some("d"), "e"),
+                    mora("ス", Some("s"), "U"),
+                ],
+                1,
+                None,
+                false,
+            ),
+        ]
+    )]
+    #[case(
+        "１、１０００、１００万、１億？",
+        &[
+            "xx^xx-sil+i=ch/A:xx+xx+xx/B:xx-xx_xx/C:xx_xx+xx/D:05+xx_xx/E:xx_xx!xx_xx-xx/F:xx_xx#xx_xx@xx_xx|xx_xx/G:2_2%0_xx_xx/H:xx_xx/I:xx-xx@xx+xx&xx-xx|xx+xx/J:1_2/K:4+4-12",
+            "xx^sil-i+ch=i/A:-1+1+2/B:xx-xx_xx/C:05_xx+xx/D:05+xx_xx/E:xx_xx!xx_xx-xx/F:2_2#0_xx@1_1|1_2/G:2_1%0_xx_0/H:xx_xx/I:1-2@1+4&1-4|1+12/J:1_2/K:4+4-12",
+            "sil^i-ch+i=pau/A:0+2+1/B:xx-xx_xx/C:05_xx+xx/D:05+xx_xx/E:xx_xx!xx_xx-xx/F:2_2#0_xx@1_1|1_2/G:2_1%0_xx_0/H:xx_xx/I:1-2@1+4&1-4|1+12/J:1_2/K:4+4-12",
+            "i^ch-i+pau=s/A:0+2+1/B:xx-xx_xx/C:05_xx+xx/D:05+xx_xx/E:xx_xx!xx_xx-xx/F:2_2#0_xx@1_1|1_2/G:2_1%0_xx_0/H:xx_xx/I:1-2@1+4&1-4|1+12/J:1_2/K:4+4-12",
+            "ch^i-pau+s=e/A:xx+xx+xx/B:05-xx_xx/C:xx_xx+xx/D:05+xx_xx/E:2_2!0_xx-xx/F:xx_xx#xx_xx@xx_xx|xx_xx/G:2_1%0_xx_xx/H:1_2/I:xx-xx@xx+xx&xx-xx|xx+xx/J:1_2/K:4+4-12",
+            "i^pau-s+e=N/A:0+1+2/B:05-xx_xx/C:05_xx+xx/D:05+xx_xx/E:2_2!0_xx-0/F:2_1#0_xx@1_1|1_2/G:4_3%0_xx_0/H:1_2/I:1-2@2+3&2-3|3+10/J:1_4/K:4+4-12",
+            "pau^s-e+N=pau/A:0+1+2/B:05-xx_xx/C:05_xx+xx/D:05+xx_xx/E:2_2!0_xx-0/F:2_1#0_xx@1_1|1_2/G:4_3%0_xx_0/H:1_2/I:1-2@2+3&2-3|3+10/J:1_4/K:4+4-12",
+            "s^e-N+pau=hy/A:1+2+1/B:05-xx_xx/C:05_xx+xx/D:05+xx_xx/E:2_2!0_xx-0/F:2_1#0_xx@1_1|1_2/G:4_3%0_xx_0/H:1_2/I:1-2@2+3&2-3|3+10/J:1_4/K:4+4-12",
+            "e^N-pau+hy=a/A:xx+xx+xx/B:05-xx_xx/C:xx_xx+xx/D:05+xx_xx/E:2_1!0_xx-xx/F:xx_xx#xx_xx@xx_xx|xx_xx/G:4_3%0_xx_xx/H:1_2/I:xx-xx@xx+xx&xx-xx|xx+xx/J:1_4/K:4+4-12",
+            "N^pau-hy+a=k/A:-2+1+4/B:05-xx_xx/C:05_xx+xx/D:05+xx_xx/E:2_1!0_xx-0/F:4_3#0_xx@1_1|1_4/G:4_2%1_xx_0/H:1_2/I:1-4@3+2&3-2|5+8/J:1_4/K:4+4-12",
+            "pau^hy-a+k=u/A:-2+1+4/B:05-xx_xx/C:05_xx+xx/D:05+xx_xx/E:2_1!0_xx-0/F:4_3#0_xx@1_1|1_4/G:4_2%1_xx_0/H:1_2/I:1-4@3+2&3-2|5+8/J:1_4/K:4+4-12",
+            "hy^a-k+u=m/A:-1+2+3/B:05-xx_xx/C:05_xx+xx/D:05+xx_xx/E:2_1!0_xx-0/F:4_3#0_xx@1_1|1_4/G:4_2%1_xx_0/H:1_2/I:1-4@3+2&3-2|5+8/J:1_4/K:4+4-12",
+            "a^k-u+m=a/A:-1+2+3/B:05-xx_xx/C:05_xx+xx/D:05+xx_xx/E:2_1!0_xx-0/F:4_3#0_xx@1_1|1_4/G:4_2%1_xx_0/H:1_2/I:1-4@3+2&3-2|5+8/J:1_4/K:4+4-12",
+            "k^u-m+a=N/A:0+3+2/B:05-xx_xx/C:05_xx+xx/D:05+xx_xx/E:2_1!0_xx-0/F:4_3#0_xx@1_1|1_4/G:4_2%1_xx_0/H:1_2/I:1-4@3+2&3-2|5+8/J:1_4/K:4+4-12",
+            "u^m-a+N=pau/A:0+3+2/B:05-xx_xx/C:05_xx+xx/D:05+xx_xx/E:2_1!0_xx-0/F:4_3#0_xx@1_1|1_4/G:4_2%1_xx_0/H:1_2/I:1-4@3+2&3-2|5+8/J:1_4/K:4+4-12",
+            "m^a-N+pau=i/A:1+4+1/B:05-xx_xx/C:05_xx+xx/D:05+xx_xx/E:2_1!0_xx-0/F:4_3#0_xx@1_1|1_4/G:4_2%1_xx_0/H:1_2/I:1-4@3+2&3-2|5+8/J:1_4/K:4+4-12",
+            "a^N-pau+i=ch/A:xx+xx+xx/B:05-xx_xx/C:xx_xx+xx/D:05+xx_xx/E:4_3!0_xx-xx/F:xx_xx#xx_xx@xx_xx|xx_xx/G:4_2%1_xx_xx/H:1_4/I:xx-xx@xx+xx&xx-xx|xx+xx/J:1_4/K:4+4-12",
+            "N^pau-i+ch=i/A:-1+1+4/B:05-xx_xx/C:05_xx+xx/D:05+xx_xx/E:4_3!0_xx-0/F:4_2#1_xx@1_1|1_4/G:xx_xx%xx_xx_xx/H:1_4/I:1-4@4+1&4-1|9+4/J:xx_xx/K:4+4-12",
+            "pau^i-ch+i=o/A:0+2+3/B:05-xx_xx/C:05_xx+xx/D:05+xx_xx/E:4_3!0_xx-0/F:4_2#1_xx@1_1|1_4/G:xx_xx%xx_xx_xx/H:1_4/I:1-4@4+1&4-1|9+4/J:xx_xx/K:4+4-12",
+            "i^ch-i+o=k/A:0+2+3/B:05-xx_xx/C:05_xx+xx/D:05+xx_xx/E:4_3!0_xx-0/F:4_2#1_xx@1_1|1_4/G:xx_xx%xx_xx_xx/H:1_4/I:1-4@4+1&4-1|9+4/J:xx_xx/K:4+4-12",
+            "ch^i-o+k=u/A:1+3+2/B:05-xx_xx/C:05_xx+xx/D:xx+xx_xx/E:4_3!0_xx-0/F:4_2#1_xx@1_1|1_4/G:xx_xx%xx_xx_xx/H:1_4/I:1-4@4+1&4-1|9+4/J:xx_xx/K:4+4-12",
+            "i^o-k+u=sil/A:2+4+1/B:05-xx_xx/C:05_xx+xx/D:xx+xx_xx/E:4_3!0_xx-0/F:4_2#1_xx@1_1|1_4/G:xx_xx%xx_xx_xx/H:1_4/I:1-4@4+1&4-1|9+4/J:xx_xx/K:4+4-12",
+            "o^k-u+sil=xx/A:2+4+1/B:05-xx_xx/C:05_xx+xx/D:xx+xx_xx/E:4_3!0_xx-0/F:4_2#1_xx@1_1|1_4/G:xx_xx%xx_xx_xx/H:1_4/I:1-4@4+1&4-1|9+4/J:xx_xx/K:4+4-12",
+            "k^u-sil+xx=xx/A:xx+xx+xx/B:05-xx_xx/C:xx_xx+xx/D:xx+xx_xx/E:4_2!1_xx-xx/F:xx_xx#xx_xx@xx_xx|xx_xx/G:xx_xx%xx_xx_xx/H:1_4/I:xx-xx@xx+xx&xx-xx|xx+xx/J:xx_xx/K:4+4-12",
+        ],
+        &[
+            AccentPhraseModel::new(
+                vec![
+                    mora("イ", None, "i"),
+                    mora("チ", Some("ch"), "i"),
+                ],
+                2,
+                Some(mora("、", None, "pau")),
+                false,
+            ),
+            AccentPhraseModel::new(
+                vec![
+                    mora("セ", Some("s"), "e"),
+                    mora("ン", None, "N"),
+                ],
+                1,
+                Some(mora("、", None, "pau")),
+                false,
+            ),
+            AccentPhraseModel::new(
+                vec![
+                    mora("ヒャ", Some("hy"), "a"),
+                    mora("ク", Some("k"), "u"),
+                    mora("マ", Some("m"), "a"),
+                    mora("ン", None, "N"),
+                ],
+                3,
+                Some(mora("、", None, "pau")),
+                false,
+            ),
+            AccentPhraseModel::new(
+                vec![
+                    mora("イ", None, "i"),
+                    mora("チ", Some("ch"), "i"),
+                    mora("オ", None, "o"),
+                    mora("ク", Some("k"), "u"),
+                ],
+                2,
+                None,
+                true,
+            ),
+        ]
+    )]
+    #[case(
+        "クヮルテット。あーあ、。",
+        &[
+            "xx^xx-sil+kw=a/A:xx+xx+xx/B:xx-xx_xx/C:xx_xx+xx/D:02+xx_xx/E:xx_xx!xx_xx-xx/F:xx_xx#xx_xx@xx_xx|xx_xx/G:5_3%0_xx_xx/H:xx_xx/I:xx-xx@xx+xx&xx-xx|xx+xx/J:1_5/K:2+3-8",
+            "xx^sil-kw+a=r/A:-2+1+5/B:xx-xx_xx/C:02_xx+xx/D:09+xx_xx/E:xx_xx!xx_xx-xx/F:5_3#0_xx@1_1|1_5/G:2_1%0_xx_0/H:xx_xx/I:1-5@1+2&1-3|1+8/J:2_3/K:2+3-8",
+            "sil^kw-a+r=u/A:-2+1+5/B:xx-xx_xx/C:02_xx+xx/D:09+xx_xx/E:xx_xx!xx_xx-xx/F:5_3#0_xx@1_1|1_5/G:2_1%0_xx_0/H:xx_xx/I:1-5@1+2&1-3|1+8/J:2_3/K:2+3-8",
+            "kw^a-r+u=t/A:-1+2+4/B:xx-xx_xx/C:02_xx+xx/D:09+xx_xx/E:xx_xx!xx_xx-xx/F:5_3#0_xx@1_1|1_5/G:2_1%0_xx_0/H:xx_xx/I:1-5@1+2&1-3|1+8/J:2_3/K:2+3-8",
+            "a^r-u+t=e/A:-1+2+4/B:xx-xx_xx/C:02_xx+xx/D:09+xx_xx/E:xx_xx!xx_xx-xx/F:5_3#0_xx@1_1|1_5/G:2_1%0_xx_0/H:xx_xx/I:1-5@1+2&1-3|1+8/J:2_3/K:2+3-8",
+            "r^u-t+e=cl/A:0+3+3/B:xx-xx_xx/C:02_xx+xx/D:09+xx_xx/E:xx_xx!xx_xx-xx/F:5_3#0_xx@1_1|1_5/G:2_1%0_xx_0/H:xx_xx/I:1-5@1+2&1-3|1+8/J:2_3/K:2+3-8",
+            "u^t-e+cl=t/A:0+3+3/B:xx-xx_xx/C:02_xx+xx/D:09+xx_xx/E:xx_xx!xx_xx-xx/F:5_3#0_xx@1_1|1_5/G:2_1%0_xx_0/H:xx_xx/I:1-5@1+2&1-3|1+8/J:2_3/K:2+3-8",
+            "t^e-cl+t=o/A:1+4+2/B:xx-xx_xx/C:02_xx+xx/D:09+xx_xx/E:xx_xx!xx_xx-xx/F:5_3#0_xx@1_1|1_5/G:2_1%0_xx_0/H:xx_xx/I:1-5@1+2&1-3|1+8/J:2_3/K:2+3-8",
+            "e^cl-t+o=pau/A:2+5+1/B:xx-xx_xx/C:02_xx+xx/D:09+xx_xx/E:xx_xx!xx_xx-xx/F:5_3#0_xx@1_1|1_5/G:2_1%0_xx_0/H:xx_xx/I:1-5@1+2&1-3|1+8/J:2_3/K:2+3-8",
+            "cl^t-o+pau=a/A:2+5+1/B:xx-xx_xx/C:02_xx+xx/D:09+xx_xx/E:xx_xx!xx_xx-xx/F:5_3#0_xx@1_1|1_5/G:2_1%0_xx_0/H:xx_xx/I:1-5@1+2&1-3|1+8/J:2_3/K:2+3-8",
+            "t^o-pau+a=a/A:xx+xx+xx/B:02-xx_xx/C:xx_xx+xx/D:09+xx_xx/E:5_3!0_xx-xx/F:xx_xx#xx_xx@xx_xx|xx_xx/G:2_1%0_xx_xx/H:1_5/I:xx-xx@xx+xx&xx-xx|xx+xx/J:2_3/K:2+3-8",
+            "o^pau-a+a=a/A:0+1+2/B:02-xx_xx/C:09_xx+xx/D:09+xx_xx/E:5_3!0_xx-0/F:2_1#0_xx@1_2|1_3/G:1_1%0_xx_1/H:1_5/I:2-3@2+1&2-2|6+3/J:xx_xx/K:2+3-8",
+            "pau^a-a+a=sil/A:1+2+1/B:02-xx_xx/C:09_xx+xx/D:09+xx_xx/E:5_3!0_xx-0/F:2_1#0_xx@1_2|1_3/G:1_1%0_xx_1/H:1_5/I:2-3@2+1&2-2|6+3/J:xx_xx/K:2+3-8",
+            "a^a-a+sil=xx/A:0+1+1/B:09-xx_xx/C:09_xx+xx/D:xx+xx_xx/E:2_1!0_xx-1/F:1_1#0_xx@2_1|3_1/G:xx_xx%xx_xx_xx/H:1_5/I:2-3@2+1&2-2|6+3/J:xx_xx/K:2+3-8",
+            "a^a-sil+xx=xx/A:xx+xx+xx/B:09-xx_xx/C:xx_xx+xx/D:xx+xx_xx/E:1_1!0_xx-xx/F:xx_xx#xx_xx@xx_xx|xx_xx/G:xx_xx%xx_xx_xx/H:2_3/I:xx-xx@xx+xx&xx-xx|xx+xx/J:xx_xx/K:2+3-8",
+        ],
+        &[
+            AccentPhraseModel::new(
+                vec![
+                    mora("クヮ", Some("kw"), "a"),
+                    mora("ル", Some("r"), "u"),
+                    mora("テ", Some("t"), "e"),
+                    mora("ッ", None, "cl"),
+                    mora("ト", Some("t"), "o"),
+                ],
+                3,
+                Some(mora("、", None, "pau")),
+                false,
+            ),
+            AccentPhraseModel::new(
+                vec![
+                    mora("ア", None, "a"),
+                    mora("ア", None, "a"),
+                ],
+                1,
+                None,
+                false,
+            ),
+            AccentPhraseModel::new(
+                vec![mora("ア", None, "a")],
+                1,
+                None,
+                false,
+            ),
+        ]
+    )]
+    fn label_cases(
+        #[case] text: &str,
+        #[case] labels: &[&str],
+        #[case] accent_phrase: &[AccentPhraseModel],
+    ) {
     }
 
-    #[allow(dead_code)]
-    pub fn phonemes(&self) -> Vec<Phoneme> {
-        // TODO:実装が中途半端なのであとでちゃんと実装する必要があるらしい
-        // https://github.com/VOICEVOX/voicevox_core/pull/174#discussion_r919982651
-        let mut phonemes = Vec::with_capacity(self.breath_groups.len());
-
-        for i in 0..self.pauses().len() {
-            phonemes.push(self.pauses().get(i).unwrap().clone());
-            if i < self.pauses().len() - 1 {
-                let p = self.breath_groups().get(i).unwrap().phonemes();
-                phonemes.extend(p);
-            }
-        }
-        phonemes
+    #[apply(label_cases)]
+    #[tokio::test]
+    async fn open_jtalk(text: &str, labels: &[&str], _accent_phrase: &[AccentPhraseModel]) {
+        let open_jtalk = crate::tokio::OpenJtalk::new(OPEN_JTALK_DIC_DIR)
+            .await
+            .unwrap();
+        assert_eq!(&open_jtalk.extract_fullcontext(text).unwrap(), labels);
     }
 
-    #[allow(dead_code)]
-    pub fn labels(&self) -> Vec<String> {
-        self.phonemes().iter().map(|p| p.label().clone()).collect()
+    #[apply(label_cases)]
+    fn parse_labels(_text: &str, labels: &[&str], accent_phrase: &[AccentPhraseModel]) {
+        let parsed_labels = labels
+            .iter()
+            .map(|s| Label::from_str(s).unwrap())
+            .collect::<Vec<_>>();
+
+        assert_eq!(
+            &generate_accent_phrases(&parsed_labels).unwrap(),
+            accent_phrase
+        );
     }
 
-    pub(crate) fn extract_full_context_label(
-        open_jtalk: &impl FullcontextExtractor,
-        text: impl AsRef<str>,
-    ) -> Result<Self> {
-        let labels = open_jtalk
-            .extract_fullcontext(text.as_ref())
-            .map_err(|source| FullContextLabelError {
-                context: ErrorKind::OpenJtalk,
-                source: Some(source),
-            })?;
-
-        labels
-            .into_iter()
-            .map(Phoneme::from_label)
-            .collect::<std::result::Result<Vec<_>, _>>()
-            .and_then(Self::from_phonemes)
-            .map_err(|context| FullContextLabelError {
-                context,
-                source: None,
-            })
+    #[apply(label_cases)]
+    #[tokio::test]
+    async fn extract_fullcontext(
+        text: &str,
+        _labels: &[&str],
+        accent_phrase: &[AccentPhraseModel],
+    ) {
+        let open_jtalk = crate::tokio::OpenJtalk::new(OPEN_JTALK_DIC_DIR)
+            .await
+            .unwrap();
+        assert_eq!(
+            &extract_full_context_label(&open_jtalk, text).unwrap(),
+            accent_phrase
+        );
     }
 }
diff --git a/crates/voicevox_core/src/engine/kana_parser.rs b/crates/voicevox_core/src/engine/kana_parser.rs
index fe10c5b3f..eaa3d93c1 100644
--- a/crates/voicevox_core/src/engine/kana_parser.rs
+++ b/crates/voicevox_core/src/engine/kana_parser.rs
@@ -163,7 +163,7 @@ pub(crate) fn parse_kana(text: &str) -> KanaParseResult<Vec<AccentPhraseModel>>
     Ok(parsed_result)
 }
 
-pub fn create_kana(accent_phrases: &[AccentPhraseModel]) -> String {
+pub(crate) fn create_kana(accent_phrases: &[AccentPhraseModel]) -> String {
     let mut text = String::new();
     for phrase in accent_phrases {
         let moras = phrase.moras();
diff --git a/crates/voicevox_core/src/engine/mod.rs b/crates/voicevox_core/src/engine/mod.rs
index 1c7422e76..95fe3d562 100644
--- a/crates/voicevox_core/src/engine/mod.rs
+++ b/crates/voicevox_core/src/engine/mod.rs
@@ -6,7 +6,9 @@ mod mora_list;
 pub(crate) mod open_jtalk;
 
 pub(crate) use self::acoustic_feature_extractor::OjtPhoneme;
-pub(crate) use self::full_context_label::{FullContextLabelError, Utterance};
+pub(crate) use self::full_context_label::{
+    extract_full_context_label, mora_to_text, FullContextLabelError,
+};
 pub(crate) use self::kana_parser::{create_kana, parse_kana, KanaParseError};
 pub use self::model::{AccentPhraseModel, AudioQueryModel, MoraModel};
 pub(crate) use self::mora_list::mora2text;
diff --git a/crates/voicevox_core/src/engine/model.rs b/crates/voicevox_core/src/engine/model.rs
index 77adbebe7..de0f388f9 100644
--- a/crates/voicevox_core/src/engine/model.rs
+++ b/crates/voicevox_core/src/engine/model.rs
@@ -5,7 +5,7 @@ use serde::{Deserialize, Serialize};
 /* 各フィールドのjsonフィールド名はsnake_caseとする*/
 
 /// モーラ（子音＋母音）ごとの情報。
-#[derive(Clone, Debug, new, Getters, Deserialize, Serialize)]
+#[derive(Clone, Debug, new, Getters, Deserialize, Serialize, PartialEq)]
 pub struct MoraModel {
     /// 文字。
     text: String,
@@ -22,7 +22,7 @@ pub struct MoraModel {
 }
 
 /// AccentPhrase (アクセント句ごとの情報)。
-#[derive(Clone, Debug, new, Getters, Deserialize, Serialize)]
+#[derive(Clone, Debug, new, Getters, Deserialize, Serialize, PartialEq)]
 pub struct AccentPhraseModel {
     /// モーラの配列。
     moras: Vec<MoraModel>,
diff --git a/crates/voicevox_core/src/engine/mora_list.rs b/crates/voicevox_core/src/engine/mora_list.rs
index 9c2c8d188..6e4b8ba2b 100644
--- a/crates/voicevox_core/src/engine/mora_list.rs
+++ b/crates/voicevox_core/src/engine/mora_list.rs
@@ -186,7 +186,7 @@ pub(super) const MORA_LIST_MINIMUM: &[[&str; 3]] = &[
     ["ア", "", "a"],
 ];
 
-pub fn mora2text(mora: &str) -> &str {
+pub(crate) fn mora2text(mora: &str) -> &str {
     for &[text, consonant, vowel] in MORA_LIST_MINIMUM {
         if mora.len() >= consonant.len()
             && &mora[..consonant.len()] == consonant
diff --git a/crates/voicevox_core/src/infer/status.rs b/crates/voicevox_core/src/infer/status.rs
index 07c189e47..65e9344af 100644
--- a/crates/voicevox_core/src/infer/status.rs
+++ b/crates/voicevox_core/src/infer/status.rs
@@ -31,14 +31,14 @@ pub(crate) struct Status<R: InferenceRuntime, D: InferenceDomain> {
 }
 
 impl<R: InferenceRuntime, D: InferenceDomain> Status<R, D> {
-    pub fn new(session_options: EnumMap<D::Operation, InferenceSessionOptions>) -> Self {
+    pub(crate) fn new(session_options: EnumMap<D::Operation, InferenceSessionOptions>) -> Self {
         Self {
             loaded_models: Default::default(),
             session_options,
         }
     }
 
-    pub fn insert_model(
+    pub(crate) fn insert_model(
         &self,
         model_header: &VoiceModelHeader,
         model_bytes: &EnumMap<D::Operation, Vec<u8>>,
@@ -64,11 +64,11 @@ impl<R: InferenceRuntime, D: InferenceDomain> Status<R, D> {
         Ok(())
     }
 
-    pub fn unload_model(&self, voice_model_id: &VoiceModelId) -> Result<()> {
+    pub(crate) fn unload_model(&self, voice_model_id: &VoiceModelId) -> Result<()> {
         self.loaded_models.lock().unwrap().remove(voice_model_id)
     }
 
-    pub fn metas(&self) -> VoiceModelMeta {
+    pub(crate) fn metas(&self) -> VoiceModelMeta {
         self.loaded_models.lock().unwrap().metas()
     }
 
@@ -76,18 +76,18 @@ impl<R: InferenceRuntime, D: InferenceDomain> Status<R, D> {
         self.loaded_models.lock().unwrap().ids_for(style_id)
     }
 
-    pub fn is_loaded_model(&self, voice_model_id: &VoiceModelId) -> bool {
+    pub(crate) fn is_loaded_model(&self, voice_model_id: &VoiceModelId) -> bool {
         self.loaded_models
             .lock()
             .unwrap()
             .contains_voice_model(voice_model_id)
     }
 
-    pub fn is_loaded_model_by_style_id(&self, style_id: StyleId) -> bool {
+    pub(crate) fn is_loaded_model_by_style_id(&self, style_id: StyleId) -> bool {
         self.loaded_models.lock().unwrap().contains_style(style_id)
     }
 
-    pub fn validate_speaker_id(&self, style_id: StyleId) -> bool {
+    pub(crate) fn validate_speaker_id(&self, style_id: StyleId) -> bool {
         self.is_loaded_model_by_style_id(style_id)
     }
 
diff --git a/crates/voicevox_core/src/lib.rs b/crates/voicevox_core/src/lib.rs
index ea74c9f7c..7cb4abaf1 100644
--- a/crates/voicevox_core/src/lib.rs
+++ b/crates/voicevox_core/src/lib.rs
@@ -12,6 +12,7 @@ mod numerics;
 mod result;
 mod synthesizer;
 mod task;
+mod text_analyzer;
 mod user_dict;
 mod version;
 mod voice_model;
@@ -23,6 +24,11 @@ pub mod tokio;
 #[cfg(test)]
 mod test_util;
 
+// https://crates.io/crates/rstest_reuse#use-rstest_resuse-at-the-top-of-your-crate
+#[allow(clippy::single_component_path_imports)]
+#[cfg(test)]
+use rstest_reuse;
+
 pub use self::{
     devices::SupportedDevices,
     engine::{AccentPhraseModel, AudioQueryModel, FullcontextExtractor},
diff --git a/crates/voicevox_core/src/synthesizer.rs b/crates/voicevox_core/src/synthesizer.rs
index 202e917c7..a3c34489e 100644
--- a/crates/voicevox_core/src/synthesizer.rs
+++ b/crates/voicevox_core/src/synthesizer.rs
@@ -80,7 +80,7 @@ pub(crate) mod blocking {
     use enum_map::enum_map;
 
     use crate::{
-        engine::{self, create_kana, parse_kana, MoraModel, OjtPhoneme, Utterance},
+        engine::{create_kana, mora_to_text, MoraModel, OjtPhoneme},
         error::ErrorRepr,
         infer::{
             domain::{
@@ -92,6 +92,7 @@ pub(crate) mod blocking {
             InferenceSessionOptions,
         },
         numerics::F32Ext as _,
+        text_analyzer::{KanaAnalyzer, OpenJTalkAnalyzer, TextAnalyzer},
         AccentPhraseModel, AudioQueryModel, FullcontextExtractor, Result, StyleId,
         SupportedDevices, SynthesisOptions, VoiceModelId, VoiceModelMeta,
     };
@@ -103,7 +104,8 @@ pub(crate) mod blocking {
     /// 音声シンセサイザ。
     pub struct Synthesizer<O> {
         pub(super) status: Status<InferenceRuntimeImpl, InferenceDomainImpl>,
-        open_jtalk: O,
+        open_jtalk_analyzer: OpenJTalkAnalyzer<O>,
+        kana_analyzer: KanaAnalyzer,
         use_gpu: bool,
     }
 
@@ -176,7 +178,8 @@ pub(crate) mod blocking {
 
             return Ok(Self {
                 status,
-                open_jtalk,
+                open_jtalk_analyzer: OpenJTalkAnalyzer::new(open_jtalk),
+                kana_analyzer: KanaAnalyzer,
                 use_gpu,
             });
 
@@ -373,7 +376,7 @@ pub(crate) mod blocking {
                 let pitch = (*last_mora.pitch() + ADJUST_PITCH).min(MAX_PITCH);
 
                 MoraModel::new(
-                    mora_to_text(last_mora.vowel()),
+                    mora_to_text(None, last_mora.vowel()),
                     None,
                     None,
                     last_mora.vowel().clone(),
@@ -457,7 +460,8 @@ pub(crate) mod blocking {
             kana: &str,
             style_id: StyleId,
         ) -> Result<Vec<AccentPhraseModel>> {
-            self.replace_mora_data(&parse_kana(kana)?, style_id)
+            let accent_phrases = self.kana_analyzer.analyze(kana)?;
+            self.replace_mora_data(&accent_phrases, style_id)
         }
 
         /// AccentPhraseの配列の音高・音素長を、特定の声で生成しなおす。
@@ -743,75 +747,7 @@ pub(crate) mod blocking {
             text: &str,
             style_id: StyleId,
         ) -> Result<Vec<AccentPhraseModel>> {
-            if text.is_empty() {
-                return Ok(Vec::new());
-            }
-
-            let utterance = Utterance::extract_full_context_label(&self.open_jtalk, text)?;
-
-            let accent_phrases: Vec<AccentPhraseModel> = utterance
-                .breath_groups()
-                .iter()
-                .enumerate()
-                .fold(Vec::new(), |mut accum_vec, (i, breath_group)| {
-                    accum_vec.extend(breath_group.accent_phrases().iter().enumerate().map(
-                        |(j, accent_phrase)| {
-                            let moras = accent_phrase
-                                .moras()
-                                .iter()
-                                .map(|mora| {
-                                    let mora_text = mora
-                                        .phonemes()
-                                        .iter()
-                                        .map(|phoneme| phoneme.phoneme().to_string())
-                                        .collect::<Vec<_>>()
-                                        .join("");
-
-                                    let (consonant, consonant_length) =
-                                        if let Some(consonant) = mora.consonant() {
-                                            (Some(consonant.phoneme().to_string()), Some(0.))
-                                        } else {
-                                            (None, None)
-                                        };
-
-                                    MoraModel::new(
-                                        mora_to_text(mora_text),
-                                        consonant,
-                                        consonant_length,
-                                        mora.vowel().phoneme().into(),
-                                        0.,
-                                        0.,
-                                    )
-                                })
-                                .collect();
-
-                            let pause_mora = if i != utterance.breath_groups().len() - 1
-                                && j == breath_group.accent_phrases().len() - 1
-                            {
-                                Some(MoraModel::new(
-                                    "、".into(),
-                                    None,
-                                    None,
-                                    "pau".into(),
-                                    0.,
-                                    0.,
-                                ))
-                            } else {
-                                None
-                            };
-
-                            AccentPhraseModel::new(
-                                moras,
-                                *accent_phrase.accent(),
-                                pause_mora,
-                                *accent_phrase.is_interrogative(),
-                            )
-                        },
-                    ));
-
-                    accum_vec
-                });
-
+            let accent_phrases = self.open_jtalk_analyzer.analyze(text)?;
             self.replace_mora_data(&accent_phrases, style_id)
         }
 
@@ -1175,21 +1111,6 @@ pub(crate) mod blocking {
         (consonant_phoneme_list, vowel_phoneme_list, vowel_indexes)
     }
 
-    fn mora_to_text(mora: impl AsRef<str>) -> String {
-        let last_char = mora.as_ref().chars().last().unwrap();
-        let mora = if ['A', 'I', 'U', 'E', 'O'].contains(&last_char) {
-            format!(
-                "{}{}",
-                &mora.as_ref()[0..mora.as_ref().len() - 1],
-                last_char.to_lowercase()
-            )
-        } else {
-            mora.as_ref().to_string()
-        };
-        // もしカタカナに変換できなければ、引数で与えた文字列がそのまま返ってくる
-        engine::mora2text(&mora).to_string()
-    }
-
     impl AudioQueryModel {
         fn from_accent_phrases(accent_phrases: Vec<AccentPhraseModel>) -> Self {
             let kana = create_kana(&accent_phrases);
diff --git a/crates/voicevox_core/src/test_util.rs b/crates/voicevox_core/src/test_util.rs
index 926fe45bb..d60785246 100644
--- a/crates/voicevox_core/src/test_util.rs
+++ b/crates/voicevox_core/src/test_util.rs
@@ -2,7 +2,7 @@ use std::path::PathBuf;
 
 use crate::Result;
 
-pub async fn open_default_vvm_file() -> crate::tokio::VoiceModel {
+pub(crate) async fn open_default_vvm_file() -> crate::tokio::VoiceModel {
     crate::tokio::VoiceModel::from_path(
         ::test_util::convert_zip_vvm(
             PathBuf::from(env!("CARGO_WORKSPACE_DIR"))
diff --git a/crates/voicevox_core/src/text_analyzer.rs b/crates/voicevox_core/src/text_analyzer.rs
new file mode 100644
index 000000000..8540f26e0
--- /dev/null
+++ b/crates/voicevox_core/src/text_analyzer.rs
@@ -0,0 +1,40 @@
+use crate::{
+    engine::{extract_full_context_label, parse_kana},
+    AccentPhraseModel, FullcontextExtractor, Result,
+};
+
+pub(crate) trait TextAnalyzer {
+    fn analyze(&self, text: &str) -> Result<Vec<AccentPhraseModel>>;
+}
+
+/// AquesTalk風記法からAccentPhraseの配列を生成するTextAnalyzer
+#[derive(Clone)]
+pub(crate) struct KanaAnalyzer;
+
+impl TextAnalyzer for KanaAnalyzer {
+    fn analyze(&self, text: &str) -> Result<Vec<AccentPhraseModel>> {
+        if text.is_empty() {
+            return Ok(Vec::new());
+        }
+        Ok(parse_kana(text)?)
+    }
+}
+
+/// OpenJtalkからAccentPhraseの配列を生成するTextAnalyzer
+#[derive(Clone)]
+pub(crate) struct OpenJTalkAnalyzer<O>(O);
+
+impl<O> OpenJTalkAnalyzer<O> {
+    pub(crate) fn new(open_jtalk: O) -> Self {
+        Self(open_jtalk)
+    }
+}
+
+impl<O: FullcontextExtractor> TextAnalyzer for OpenJTalkAnalyzer<O> {
+    fn analyze(&self, text: &str) -> Result<Vec<AccentPhraseModel>> {
+        if text.is_empty() {
+            return Ok(Vec::new());
+        }
+        Ok(extract_full_context_label(&self.0, text)?)
+    }
+}
diff --git a/crates/voicevox_core/src/user_dict/part_of_speech_data.rs b/crates/voicevox_core/src/user_dict/part_of_speech_data.rs
index 76e36e389..b7bc95440 100644
--- a/crates/voicevox_core/src/user_dict/part_of_speech_data.rs
+++ b/crates/voicevox_core/src/user_dict/part_of_speech_data.rs
@@ -1,37 +1,36 @@
-use derive_getters::Getters;
 use once_cell::sync::Lazy;
 use std::collections::HashMap;
 
 use crate::UserDictWordType;
 
 /// 最小の優先度。
-pub static MIN_PRIORITY: u32 = 0;
+pub(super) static MIN_PRIORITY: u32 = 0;
 /// 最大の優先度。
-pub static MAX_PRIORITY: u32 = 10;
+pub(super) static MAX_PRIORITY: u32 = 10;
 
 /// 品詞ごとの情報。
-#[derive(Debug, Getters)]
-pub struct PartOfSpeechDetail {
+#[derive(Debug)]
+pub(super) struct PartOfSpeechDetail {
     /// 品詞。
-    pub part_of_speech: &'static str,
+    pub(super) part_of_speech: &'static str,
     /// 品詞細分類1。
-    pub part_of_speech_detail_1: &'static str,
+    pub(super) part_of_speech_detail_1: &'static str,
     /// 品詞細分類2。
-    pub part_of_speech_detail_2: &'static str,
+    pub(super) part_of_speech_detail_2: &'static str,
     /// 品詞細分類3。
-    pub part_of_speech_detail_3: &'static str,
+    pub(super) part_of_speech_detail_3: &'static str,
     /// 文脈IDは辞書の左・右文脈IDのこと。
     ///
     /// 参考: <https://github.com/VOICEVOX/open_jtalk/blob/427cfd761b78efb6094bea3c5bb8c968f0d711ab/src/mecab-naist-jdic/_left-id.def>
-    pub context_id: i32,
+    pub(super) context_id: i32,
     /// コストのパーセンタイル。
-    pub cost_candidates: Vec<i32>,
+    cost_candidates: Vec<i32>,
     /// アクセント結合規則の一覧。
-    pub accent_associative_rules: Vec<&'static str>,
+    _accent_associative_rules: Vec<&'static str>, // unused for now
 }
 
 // 元データ： https://github.com/VOICEVOX/voicevox_engine/blob/master/voicevox_engine/part_of_speech_data.py
-pub static PART_OF_SPEECH_DETAIL: Lazy<HashMap<UserDictWordType, PartOfSpeechDetail>> =
+pub(super) static PART_OF_SPEECH_DETAIL: Lazy<HashMap<UserDictWordType, PartOfSpeechDetail>> =
     Lazy::new(|| {
         HashMap::from_iter([
             (
@@ -45,7 +44,7 @@ pub static PART_OF_SPEECH_DETAIL: Lazy<HashMap<UserDictWordType, PartOfSpeechDet
                     cost_candidates: vec![
                         -988, 3488, 4768, 6048, 7328, 8609, 8734, 8859, 8984, 9110, 14176,
                     ],
-                    accent_associative_rules: vec!["*", "C1", "C2", "C3", "C4", "C5"],
+                    _accent_associative_rules: vec!["*", "C1", "C2", "C3", "C4", "C5"],
                 },
             ),
             (
@@ -59,7 +58,7 @@ pub static PART_OF_SPEECH_DETAIL: Lazy<HashMap<UserDictWordType, PartOfSpeechDet
                     cost_candidates: vec![
                         -4445, 49, 1473, 2897, 4321, 5746, 6554, 7362, 8170, 8979, 15001,
                     ],
-                    accent_associative_rules: vec!["*", "C1", "C2", "C3", "C4", "C5"],
+                    _accent_associative_rules: vec!["*", "C1", "C2", "C3", "C4", "C5"],
                 },
             ),
             (
@@ -73,7 +72,7 @@ pub static PART_OF_SPEECH_DETAIL: Lazy<HashMap<UserDictWordType, PartOfSpeechDet
                     cost_candidates: vec![
                         3100, 6160, 6360, 6561, 6761, 6962, 7414, 7866, 8318, 8771, 13433,
                     ],
-                    accent_associative_rules: vec!["*"],
+                    _accent_associative_rules: vec!["*"],
                 },
             ),
             (
@@ -87,7 +86,7 @@ pub static PART_OF_SPEECH_DETAIL: Lazy<HashMap<UserDictWordType, PartOfSpeechDet
                     cost_candidates: vec![
                         1527, 3266, 3561, 3857, 4153, 4449, 5149, 5849, 6549, 7250, 10001,
                     ],
-                    accent_associative_rules: vec!["*"],
+                    _accent_associative_rules: vec!["*"],
                 },
             ),
             (
@@ -101,7 +100,7 @@ pub static PART_OF_SPEECH_DETAIL: Lazy<HashMap<UserDictWordType, PartOfSpeechDet
                     cost_candidates: vec![
                         4399, 5373, 6041, 6710, 7378, 8047, 9440, 10834, 12228, 13622, 15847,
                     ],
-                    accent_associative_rules: vec!["*", "C1", "C2", "C3", "C4", "C5"],
+                    _accent_associative_rules: vec!["*", "C1", "C2", "C3", "C4", "C5"],
                 },
             ),
         ])
@@ -115,7 +114,7 @@ fn search_cost_candidates(context_id: i32) -> &'static [i32] {
         .cost_candidates
 }
 
-pub fn priority2cost(context_id: i32, priority: u32) -> i32 {
+pub(super) fn priority2cost(context_id: i32, priority: u32) -> i32 {
     let cost_candidates = search_cost_candidates(context_id);
     cost_candidates[(MAX_PRIORITY - priority) as usize]
 }
diff --git a/crates/voicevox_core/src/user_dict/word.rs b/crates/voicevox_core/src/user_dict/word.rs
index f2abc905d..7ed98a949 100644
--- a/crates/voicevox_core/src/user_dict/word.rs
+++ b/crates/voicevox_core/src/user_dict/word.rs
@@ -219,7 +219,7 @@ pub enum UserDictWordType {
 }
 
 impl UserDictWord {
-    pub fn to_mecab_format(&self) -> String {
+    pub(super) fn to_mecab_format(&self) -> String {
         let pos = PART_OF_SPEECH_DETAIL.get(&self.word_type).unwrap();
         format!(
             "{},{},{},{},{},{},{},{},{},{},{},{},{},{}/{},{}",
diff --git a/crates/voicevox_core/src/voice_model.rs b/crates/voicevox_core/src/voice_model.rs
index fc8f4d20f..76f692eaa 100644
--- a/crates/voicevox_core/src/voice_model.rs
+++ b/crates/voicevox_core/src/voice_model.rs
@@ -1,3 +1,7 @@
+//! 音声モデル（ VVM ファイル）。
+//!
+//! VVM ファイルの定義と形式は[ドキュメント](../../../docs/vvm.md)を参照。
+
 use derive_getters::Getters;
 use derive_new::new;
 use serde::Deserialize;
@@ -287,42 +291,43 @@ pub(crate) mod tokio {
     }
 
     #[derive(new)]
-    struct AsyncVvmEntryReader {
-        reader: async_zip::read::fs::ZipFileReader,
+    struct AsyncVvmEntryReader<'a> {
+        path: &'a Path,
+        reader: async_zip::base::read::mem::ZipFileReader,
         entry_map: HashMap<String, AsyncVvmEntry>,
     }
 
-    impl AsyncVvmEntryReader {
-        async fn open(path: &Path) -> LoadModelResult<Self> {
-            let reader = async_zip::read::fs::ZipFileReader::new(path)
-                .await
-                .map_err(|source| LoadModelError {
-                    path: path.to_owned(),
-                    context: LoadModelErrorKind::OpenZipFile,
-                    source: Some(source.into()),
-                })?;
+    impl<'a> AsyncVvmEntryReader<'a> {
+        async fn open(path: &'a Path) -> LoadModelResult<Self> {
+            let reader = async {
+                let file = fs_err::tokio::read(path).await?;
+                async_zip::base::read::mem::ZipFileReader::new(file).await
+            }
+            .await
+            .map_err(|source| LoadModelError {
+                path: path.to_owned(),
+                context: LoadModelErrorKind::OpenZipFile,
+                source: Some(source.into()),
+            })?;
             let entry_map: HashMap<_, _> = reader
                 .file()
                 .entries()
                 .iter()
-                .filter(|e| !e.entry().dir())
-                .enumerate()
-                .map(|(i, e)| {
-                    (
-                        e.entry().filename().to_string(),
-                        AsyncVvmEntry {
-                            index: i,
-                            entry: e.entry().clone(),
-                        },
-                    )
+                .flat_map(|e| {
+                    // 非UTF-8のファイルを利用することはないため、無視する
+                    let filename = e.filename().as_str().ok()?;
+                    (!e.dir().ok()?).then_some(())?;
+                    Some((filename.to_owned(), (**e).clone()))
                 })
+                .enumerate()
+                .map(|(i, (filename, entry))| (filename, AsyncVvmEntry { index: i, entry }))
                 .collect();
-            Ok(AsyncVvmEntryReader::new(reader, entry_map))
+            Ok(AsyncVvmEntryReader::new(path, reader, entry_map))
         }
         async fn read_vvm_json<T: DeserializeOwned>(&self, filename: &str) -> LoadModelResult<T> {
             let bytes = self.read_vvm_entry(filename).await?;
             serde_json::from_slice(&bytes).map_err(|source| LoadModelError {
-                path: self.reader.path().to_owned(),
+                path: self.path.to_owned(),
                 context: LoadModelErrorKind::ReadZipEntry {
                     filename: filename.to_owned(),
                 },
@@ -336,16 +341,14 @@ pub(crate) mod tokio {
                     .entry_map
                     .get(filename)
                     .ok_or_else(|| io::Error::from(io::ErrorKind::NotFound))?;
-                let mut manifest_reader = self.reader.entry(me.index).await?;
+                let mut manifest_reader = self.reader.reader_with_entry(me.index).await?;
                 let mut buf = Vec::with_capacity(me.entry.uncompressed_size() as usize);
-                manifest_reader
-                    .read_to_end_checked(&mut buf, &me.entry)
-                    .await?;
+                manifest_reader.read_to_end_checked(&mut buf).await?;
                 Ok::<_, anyhow::Error>(buf)
             }
             .await
             .map_err(|source| LoadModelError {
-                path: self.reader.path().to_owned(),
+                path: self.path.to_owned(),
                 context: LoadModelErrorKind::ReadZipEntry {
                     filename: filename.to_owned(),
                 },
diff --git a/crates/voicevox_core_c_api/src/helpers.rs b/crates/voicevox_core_c_api/src/helpers.rs
index d69641c34..74177db9c 100644
--- a/crates/voicevox_core_c_api/src/helpers.rs
+++ b/crates/voicevox_core_c_api/src/helpers.rs
@@ -63,7 +63,7 @@ pub(crate) fn into_result_code_with_error(result: CApiResult<()>) -> VoicevoxRes
 pub(crate) type CApiResult<T> = std::result::Result<T, CApiError>;
 
 #[derive(Error, Debug)]
-pub enum CApiError {
+pub(crate) enum CApiError {
     #[error("{0}")]
     RustApi(#[from] voicevox_core::Error),
     #[error("UTF-8として不正な入力です")]
diff --git a/crates/voicevox_core_c_api/src/lib.rs b/crates/voicevox_core_c_api/src/lib.rs
index 61127ff81..fbb0bf6bf 100644
--- a/crates/voicevox_core_c_api/src/lib.rs
+++ b/crates/voicevox_core_c_api/src/lib.rs
@@ -447,8 +447,10 @@ pub unsafe extern "C" fn voicevox_synthesizer_is_loaded_voice_model(
     model_id: VoicevoxVoiceModelId,
 ) -> bool {
     init_logger_once();
-    // FIXME: 不正なUTF-8文字列に対し、正式なエラーとするか黙って`false`を返す
-    let raw_model_id = ensure_utf8(unsafe { CStr::from_ptr(model_id) }).unwrap();
+    let Ok(raw_model_id) = ensure_utf8(unsafe { CStr::from_ptr(model_id) }) else {
+        // 与えられたIDがUTF-8ではない場合、それに対応する`VoicdModel`は確実に存在しない
+        return false;
+    };
     synthesizer
         .synthesizer()
         .is_loaded_voice_model(&VoiceModelId::new(raw_model_id.into()))
diff --git a/crates/voicevox_core_c_api/xcframework/Frameworks/aarch64/voicevox_core.framework/Info.plist b/crates/voicevox_core_c_api/xcframework/Frameworks/aarch64/voicevox_core.framework/Info.plist
new file mode 100644
index 000000000..f3f359d09
--- /dev/null
+++ b/crates/voicevox_core_c_api/xcframework/Frameworks/aarch64/voicevox_core.framework/Info.plist
@@ -0,0 +1,55 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
+<plist version="1.0">
+<dict>
+	<key>BuildMachineOSBuild</key>
+	<string>23B81</string>
+	<key>CFBundleDevelopmentRegion</key>
+	<string>en</string>
+	<key>CFBundleExecutable</key>
+	<string>voicevox_core</string>
+	<key>CFBundleIdentifier</key>
+	<string>jp.hiroshiba.voicevox.voicevox-core</string>
+	<key>CFBundleInfoDictionaryVersion</key>
+	<string>6.0</string>
+	<key>CFBundleName</key>
+	<string>voicevox_core</string>
+	<key>CFBundlePackageType</key>
+	<string>FMWK</string>
+	<key>CFBundleShortVersionString</key>
+	<string>1.0</string>
+	<key>CFBundleSupportedPlatforms</key>
+	<array>
+		<string>iPhoneOS</string>
+	</array>
+	<key>CFBundleVersion</key>
+	<string>1</string>
+	<key>DTCompiler</key>
+	<string>com.apple.compilers.llvm.clang.1_0</string>
+	<key>DTPlatformBuild</key>
+	<string>21C52</string>
+	<key>DTPlatformName</key>
+	<string>iphoneos</string>
+	<key>DTPlatformVersion</key>
+	<string>17.2</string>
+	<key>DTSDKBuild</key>
+	<string>21C52</string>
+	<key>DTSDKName</key>
+	<string>iphoneos17.2</string>
+	<key>DTXcode</key>
+	<string>1510</string>
+	<key>DTXcodeBuild</key>
+	<string>15C65</string>
+	<key>MinimumOSVersion</key>
+	<string>16.2</string>
+	<key>UIDeviceFamily</key>
+	<array>
+		<integer>1</integer>
+		<integer>2</integer>
+	</array>
+	<key>UIRequiredDeviceCapabilities</key>
+	<array>
+		<string>arm64</string>
+	</array>
+</dict>
+</plist>
diff --git a/crates/voicevox_core_c_api/xcframework/Frameworks/aarch64/voicevox_core.framework/Modules/module.modulemap b/crates/voicevox_core_c_api/xcframework/Frameworks/aarch64/voicevox_core.framework/Modules/module.modulemap
new file mode 100644
index 000000000..a4812ac0c
--- /dev/null
+++ b/crates/voicevox_core_c_api/xcframework/Frameworks/aarch64/voicevox_core.framework/Modules/module.modulemap
@@ -0,0 +1,6 @@
+framework module voicevox_core {
+  umbrella header "voicevox_core.h"
+  export *
+
+  module * { export * }
+}
diff --git a/crates/voicevox_core_c_api/xcframework/Frameworks/sim/voicevox_core.framework/Info.plist b/crates/voicevox_core_c_api/xcframework/Frameworks/sim/voicevox_core.framework/Info.plist
new file mode 100644
index 000000000..dea41a54d
--- /dev/null
+++ b/crates/voicevox_core_c_api/xcframework/Frameworks/sim/voicevox_core.framework/Info.plist
@@ -0,0 +1,51 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
+<plist version="1.0">
+<dict>
+	<key>BuildMachineOSBuild</key>
+	<string>23B81</string>
+	<key>CFBundleDevelopmentRegion</key>
+	<string>en</string>
+	<key>CFBundleExecutable</key>
+	<string>voicevox_core</string>
+	<key>CFBundleIdentifier</key>
+	<string>jp.hiroshiba.voicevox.voicevox-core</string>
+	<key>CFBundleInfoDictionaryVersion</key>
+	<string>6.0</string>
+	<key>CFBundleName</key>
+	<string>voicevox_core</string>
+	<key>CFBundlePackageType</key>
+	<string>FMWK</string>
+	<key>CFBundleShortVersionString</key>
+	<string>1.0</string>
+	<key>CFBundleSupportedPlatforms</key>
+	<array>
+		<string>iPhoneSimulator</string>
+	</array>
+	<key>CFBundleVersion</key>
+	<string>1</string>
+	<key>DTCompiler</key>
+	<string>com.apple.compilers.llvm.clang.1_0</string>
+	<key>DTPlatformBuild</key>
+	<string>21C52</string>
+	<key>DTPlatformName</key>
+	<string>iphonesimulator</string>
+	<key>DTPlatformVersion</key>
+	<string>17.2</string>
+	<key>DTSDKBuild</key>
+	<string>21C52</string>
+	<key>DTSDKName</key>
+	<string>iphonesimulator17.2</string>
+	<key>DTXcode</key>
+	<string>1510</string>
+	<key>DTXcodeBuild</key>
+	<string>15C65</string>
+	<key>MinimumOSVersion</key>
+	<string>16.2</string>
+	<key>UIDeviceFamily</key>
+	<array>
+		<integer>1</integer>
+		<integer>2</integer>
+	</array>
+</dict>
+</plist>
diff --git a/crates/voicevox_core_c_api/xcframework/Frameworks/sim/voicevox_core.framework/Modules/module.modulemap b/crates/voicevox_core_c_api/xcframework/Frameworks/sim/voicevox_core.framework/Modules/module.modulemap
new file mode 100644
index 000000000..a4812ac0c
--- /dev/null
+++ b/crates/voicevox_core_c_api/xcframework/Frameworks/sim/voicevox_core.framework/Modules/module.modulemap
@@ -0,0 +1,6 @@
+framework module voicevox_core {
+  umbrella header "voicevox_core.h"
+  export *
+
+  module * { export * }
+}
diff --git a/crates/voicevox_core_c_api/xcframework/Headers/README.md b/crates/voicevox_core_c_api/xcframework/Headers/README.md
deleted file mode 100644
index cd3b3f6a0..000000000
--- a/crates/voicevox_core_c_api/xcframework/Headers/README.md
+++ /dev/null
@@ -1,5 +0,0 @@
-# xcframeworkのHeadersに追加されるファイル
-## module.modulemap
-
-C言語やObjective-Cのモジュールが、ライブラリ内の公開インターフェースを表現するために使用されます。  
-これにより、外部のコードがこのライブラリを利用する際に、必要なインクルードや参照を容易にすることができます。
diff --git a/crates/voicevox_core_c_api/xcframework/Headers/module.modulemap b/crates/voicevox_core_c_api/xcframework/Headers/module.modulemap
deleted file mode 100644
index 1b1680b2d..000000000
--- a/crates/voicevox_core_c_api/xcframework/Headers/module.modulemap
+++ /dev/null
@@ -1,4 +0,0 @@
-module VoicevoxCore {
-  header "voicevox_core.h"
-  export *
-}
diff --git a/crates/voicevox_core_c_api/xcframework/README.md b/crates/voicevox_core_c_api/xcframework/README.md
new file mode 100644
index 000000000..3dda8f3a7
--- /dev/null
+++ b/crates/voicevox_core_c_api/xcframework/README.md
@@ -0,0 +1,6 @@
+# xcframeworkフォルダの内容について
+## Frameworks
+
+iOS向けの配布ライブラリXCFramework内のFrameworkを作るための雛形です。
+雛形は端末用とシミュレータ用の2種類です。
+
diff --git a/crates/voicevox_core_java_api/src/common.rs b/crates/voicevox_core_java_api/src/common.rs
index 5d1278f0c..71a60f5f3 100644
--- a/crates/voicevox_core_java_api/src/common.rs
+++ b/crates/voicevox_core_java_api/src/common.rs
@@ -1,72 +1,7 @@
 use std::{error::Error as _, iter};
 
 use derive_more::From;
-use jni::{
-    objects::{JObject, JThrowable},
-    JNIEnv,
-};
-
-// FIXME: 別ファイルに分離する
-#[no_mangle]
-extern "system" fn Java_jp_hiroshiba_voicevoxcore_Dll_00024LoggerInitializer_initLogger(
-    _: JNIEnv<'_>,
-    _: JObject<'_>,
-) {
-    if cfg!(target_os = "android") {
-        android_logger::init_once(
-            android_logger::Config::default()
-                .with_tag("VoicevoxCore")
-                .with_filter(
-                    android_logger::FilterBuilder::new()
-                        // FIXME: ortも`warn`は出すべき
-                        .parse("error,voicevox_core=info,voicevox_core_java_api=info,ort=error")
-                        .build(),
-                ),
-        );
-    } else {
-        // TODO: Android以外でのログ出力を良い感じにする。（System.Loggerを使う？）
-        use chrono::SecondsFormat;
-        use std::{
-            env, fmt,
-            io::{self, IsTerminal, Write},
-        };
-        use tracing_subscriber::{fmt::format::Writer, EnvFilter};
-
-        // FIXME: `try_init` → `init` （subscriberは他に存在しないはずなので）
-        let _ = tracing_subscriber::fmt()
-            .with_env_filter(if env::var_os(EnvFilter::DEFAULT_ENV).is_some() {
-                EnvFilter::from_default_env()
-            } else {
-                // FIXME: `c_api`じゃないし、ortも`warn`は出すべき
-                "error,voicevox_core=info,voicevox_core_c_api=info,ort=error".into()
-            })
-            .with_timer(local_time as fn(&mut Writer<'_>) -> _)
-            .with_ansi(out().is_terminal() && env_allows_ansi())
-            .with_writer(out)
-            .try_init();
-
-        fn local_time(wtr: &mut Writer<'_>) -> fmt::Result {
-            // ローカル時刻で表示はするが、そのフォーマットはtracing-subscriber本来のものに近いようにする。
-            // https://github.com/tokio-rs/tracing/blob/tracing-subscriber-0.3.16/tracing-subscriber/src/fmt/time/datetime.rs#L235-L241
-            wtr.write_str(&chrono::Local::now().to_rfc3339_opts(SecondsFormat::Micros, false))
-        }
-
-        fn out() -> impl IsTerminal + Write {
-            io::stderr()
-        }
-
-        fn env_allows_ansi() -> bool {
-            // https://docs.rs/termcolor/1.2.0/src/termcolor/lib.rs.html#245-291
-            // ただしWindowsではPowerShellっぽかったらそのまま許可する。
-            // ちゃんとやるなら`ENABLE_VIRTUAL_TERMINAL_PROCESSING`をチェックするなり、そもそも
-            // fwdansiとかでWin32の色に変換するべきだが、面倒。
-            env::var_os("TERM").map_or(
-                cfg!(windows) && env::var_os("PSModulePath").is_some(),
-                |term| term != "dumb",
-            ) && env::var_os("NO_COLOR").is_none()
-        }
-    }
-}
+use jni::{objects::JThrowable, JNIEnv};
 
 #[macro_export]
 macro_rules! object {
@@ -94,7 +29,7 @@ macro_rules! enum_object {
     };
 }
 
-pub fn throw_if_err<T, F>(mut env: JNIEnv<'_>, fallback: T, inner: F) -> T
+pub(crate) fn throw_if_err<T, F>(mut env: JNIEnv<'_>, fallback: T, inner: F) -> T
 where
     F: FnOnce(&mut JNIEnv<'_>) -> Result<T, JavaApiError>,
 {
@@ -220,7 +155,7 @@ where
 }
 
 #[derive(From, Debug)]
-pub enum JavaApiError {
+pub(crate) enum JavaApiError {
     #[from]
     RustApi(voicevox_core::Error),
 
diff --git a/crates/voicevox_core_java_api/src/lib.rs b/crates/voicevox_core_java_api/src/lib.rs
index 4fdea9cab..9615f0a94 100644
--- a/crates/voicevox_core_java_api/src/lib.rs
+++ b/crates/voicevox_core_java_api/src/lib.rs
@@ -1,5 +1,6 @@
 mod common;
 mod info;
+mod logger;
 mod open_jtalk;
 mod synthesizer;
 mod user_dict;
diff --git a/crates/voicevox_core_java_api/src/logger.rs b/crates/voicevox_core_java_api/src/logger.rs
new file mode 100644
index 000000000..30545725e
--- /dev/null
+++ b/crates/voicevox_core_java_api/src/logger.rs
@@ -0,0 +1,62 @@
+use jni::{objects::JObject, JNIEnv};
+
+#[no_mangle]
+extern "system" fn Java_jp_hiroshiba_voicevoxcore_Dll_00024LoggerInitializer_initLogger(
+    _: JNIEnv<'_>,
+    _: JObject<'_>,
+) {
+    if cfg!(target_os = "android") {
+        android_logger::init_once(
+            android_logger::Config::default()
+                .with_tag("VoicevoxCore")
+                .with_filter(
+                    android_logger::FilterBuilder::new()
+                        // FIXME: ortも`warn`は出すべき
+                        .parse("error,voicevox_core=info,voicevox_core_java_api=info,ort=error")
+                        .build(),
+                ),
+        );
+    } else {
+        // TODO: Android以外でのログ出力を良い感じにする。（System.Loggerを使う？）
+        use chrono::SecondsFormat;
+        use std::{
+            env, fmt,
+            io::{self, IsTerminal, Write},
+        };
+        use tracing_subscriber::{fmt::format::Writer, EnvFilter};
+
+        // FIXME: `try_init` → `init` （subscriberは他に存在しないはずなので）
+        let _ = tracing_subscriber::fmt()
+            .with_env_filter(if env::var_os(EnvFilter::DEFAULT_ENV).is_some() {
+                EnvFilter::from_default_env()
+            } else {
+                // FIXME: `c_api`じゃないし、ortも`warn`は出すべき
+                "error,voicevox_core=info,voicevox_core_c_api=info,ort=error".into()
+            })
+            .with_timer(local_time as fn(&mut Writer<'_>) -> _)
+            .with_ansi(out().is_terminal() && env_allows_ansi())
+            .with_writer(out)
+            .try_init();
+
+        fn local_time(wtr: &mut Writer<'_>) -> fmt::Result {
+            // ローカル時刻で表示はするが、そのフォーマットはtracing-subscriber本来のものに近いようにする。
+            // https://github.com/tokio-rs/tracing/blob/tracing-subscriber-0.3.16/tracing-subscriber/src/fmt/time/datetime.rs#L235-L241
+            wtr.write_str(&chrono::Local::now().to_rfc3339_opts(SecondsFormat::Micros, false))
+        }
+
+        fn out() -> impl IsTerminal + Write {
+            io::stderr()
+        }
+
+        fn env_allows_ansi() -> bool {
+            // https://docs.rs/termcolor/1.2.0/src/termcolor/lib.rs.html#245-291
+            // ただしWindowsではPowerShellっぽかったらそのまま許可する。
+            // ちゃんとやるなら`ENABLE_VIRTUAL_TERMINAL_PROCESSING`をチェックするなり、そもそも
+            // fwdansiとかでWin32の色に変換するべきだが、面倒。
+            env::var_os("TERM").map_or(
+                cfg!(windows) && env::var_os("PSModulePath").is_some(),
+                |term| term != "dumb",
+            ) && env::var_os("NO_COLOR").is_none()
+        }
+    }
+}
diff --git a/crates/voicevox_core_python_api/python/voicevox_core/_rust/asyncio.pyi b/crates/voicevox_core_python_api/python/voicevox_core/_rust/asyncio.pyi
index 7a6596008..d8e9f6fe2 100644
--- a/crates/voicevox_core_python_api/python/voicevox_core/_rust/asyncio.pyi
+++ b/crates/voicevox_core_python_api/python/voicevox_core/_rust/asyncio.pyi
@@ -1,4 +1,4 @@
-from pathlib import Path
+from os import PathLike
 from typing import TYPE_CHECKING, Dict, List, Literal, Union
 from uuid import UUID
 
@@ -18,7 +18,7 @@ class VoiceModel:
     音声モデル。"""
 
     @staticmethod
-    async def from_path(path: Union[Path, str]) -> VoiceModel:
+    async def from_path(path: Union[str, PathLike[str]]) -> VoiceModel:
         """
         VVMファイルから ``VoiceModel`` を生成する。
 
@@ -43,7 +43,7 @@ class OpenJtalk:
     """
 
     @staticmethod
-    async def new(open_jtalk_dict_dir: Union[Path, str]) -> "OpenJtalk":
+    async def new(open_jtalk_dict_dir: Union[str, PathLike[str]]) -> "OpenJtalk":
         """
         ``OpenJTalk`` を生成する。
 
diff --git a/crates/voicevox_core_python_api/python/voicevox_core/_rust/blocking.pyi b/crates/voicevox_core_python_api/python/voicevox_core/_rust/blocking.pyi
index 3a208fb33..5584d68bb 100644
--- a/crates/voicevox_core_python_api/python/voicevox_core/_rust/blocking.pyi
+++ b/crates/voicevox_core_python_api/python/voicevox_core/_rust/blocking.pyi
@@ -1,4 +1,4 @@
-from pathlib import Path
+from os import PathLike
 from typing import TYPE_CHECKING, Dict, List, Literal, Union
 from uuid import UUID
 
@@ -18,7 +18,7 @@ class VoiceModel:
     音声モデル。"""
 
     @staticmethod
-    def from_path(path: Union[Path, str]) -> VoiceModel:
+    def from_path(path: Union[str, PathLike[str]]) -> VoiceModel:
         """
         VVMファイルから ``VoiceModel`` を生成する。
 
@@ -47,7 +47,7 @@ class OpenJtalk:
         Open JTalkの辞書ディレクトリ。
     """
 
-    def __init__(self, open_jtalk_dict_dir: Union[Path, str]) -> None: ...
+    def __init__(self, open_jtalk_dict_dir: Union[str, PathLike[str]]) -> None: ...
     def use_user_dict(self, user_dict: UserDict) -> None:
         """
         ユーザー辞書を設定する。
diff --git a/crates/voicevox_core_python_api/src/convert.rs b/crates/voicevox_core_python_api/src/convert.rs
index 6544ce26c..3f629f90e 100644
--- a/crates/voicevox_core_python_api/src/convert.rs
+++ b/crates/voicevox_core_python_api/src/convert.rs
@@ -22,7 +22,7 @@ use crate::{
     UseUserDictError, WordNotFoundError,
 };
 
-pub fn from_acceleration_mode(ob: &PyAny) -> PyResult<AccelerationMode> {
+pub(crate) fn from_acceleration_mode(ob: &PyAny) -> PyResult<AccelerationMode> {
     let py = ob.py();
 
     let class = py.import("voicevox_core")?.getattr("AccelerationMode")?;
@@ -39,8 +39,8 @@ pub fn from_acceleration_mode(ob: &PyAny) -> PyResult<AccelerationMode> {
     }
 }
 
-// FIXME: `VoiceModel`や`UserDict`についてはこれではなく、`PathBuf::extract`を直接使うようにする
-pub fn from_utf8_path(ob: &PyAny) -> PyResult<Utf8PathBuf> {
+// FIXME: `UserDict`についてはこれではなく、`PathBuf::extract`を直接使うようにする
+pub(crate) fn from_utf8_path(ob: &PyAny) -> PyResult<Utf8PathBuf> {
     PathBuf::extract(ob)?
         .into_os_string()
         .into_string()
@@ -48,7 +48,7 @@ pub fn from_utf8_path(ob: &PyAny) -> PyResult<Utf8PathBuf> {
         .map_err(|s| PyValueError::new_err(format!("{s:?} cannot be encoded to UTF-8")))
 }
 
-pub fn from_dataclass<T: DeserializeOwned>(ob: &PyAny) -> PyResult<T> {
+pub(crate) fn from_dataclass<T: DeserializeOwned>(ob: &PyAny) -> PyResult<T> {
     let py = ob.py();
 
     let ob = py.import("dataclasses")?.call_method1("asdict", (ob,))?;
@@ -59,7 +59,7 @@ pub fn from_dataclass<T: DeserializeOwned>(ob: &PyAny) -> PyResult<T> {
     serde_json::from_str(json).into_py_value_result()
 }
 
-pub fn to_pydantic_voice_model_meta<'py>(
+pub(crate) fn to_pydantic_voice_model_meta<'py>(
     metas: &VoiceModelMeta,
     py: Python<'py>,
 ) -> PyResult<Vec<&'py PyAny>> {
@@ -74,7 +74,7 @@ pub fn to_pydantic_voice_model_meta<'py>(
         .collect::<PyResult<Vec<_>>>()
 }
 
-pub fn to_pydantic_dataclass(x: impl Serialize, class: &PyAny) -> PyResult<&PyAny> {
+pub(crate) fn to_pydantic_dataclass(x: impl Serialize, class: &PyAny) -> PyResult<&PyAny> {
     let py = class.py();
 
     let x = serde_json::to_string(&x).into_py_value_result()?;
@@ -108,7 +108,7 @@ pub(crate) fn blocking_modify_accent_phrases<'py>(
         .collect()
 }
 
-pub fn async_modify_accent_phrases<'py, Fun, Fut>(
+pub(crate) fn async_modify_accent_phrases<'py, Fun, Fut>(
     accent_phrases: &'py PyList,
     speaker_id: StyleId,
     py: Python<'py>,
@@ -145,16 +145,16 @@ where
     )
 }
 
-pub fn to_rust_uuid(ob: &PyAny) -> PyResult<Uuid> {
+pub(crate) fn to_rust_uuid(ob: &PyAny) -> PyResult<Uuid> {
     let uuid = ob.getattr("hex")?.extract::<String>()?;
     uuid.parse::<Uuid>().into_py_value_result()
 }
-pub fn to_py_uuid(py: Python<'_>, uuid: Uuid) -> PyResult<PyObject> {
+pub(crate) fn to_py_uuid(py: Python<'_>, uuid: Uuid) -> PyResult<PyObject> {
     let uuid = uuid.hyphenated().to_string();
     let uuid = py.import("uuid")?.call_method1("UUID", (uuid,))?;
     Ok(uuid.to_object(py))
 }
-pub fn to_rust_user_dict_word(ob: &PyAny) -> PyResult<voicevox_core::UserDictWord> {
+pub(crate) fn to_rust_user_dict_word(ob: &PyAny) -> PyResult<voicevox_core::UserDictWord> {
     voicevox_core::UserDictWord::new(
         ob.getattr("surface")?.extract()?,
         ob.getattr("pronunciation")?.extract()?,
@@ -164,7 +164,7 @@ pub fn to_rust_user_dict_word(ob: &PyAny) -> PyResult<voicevox_core::UserDictWor
     )
     .into_py_result(ob.py())
 }
-pub fn to_py_user_dict_word<'py>(
+pub(crate) fn to_py_user_dict_word<'py>(
     py: Python<'py>,
     word: &voicevox_core::UserDictWord,
 ) -> PyResult<&'py PyAny> {
@@ -174,14 +174,14 @@ pub fn to_py_user_dict_word<'py>(
         .downcast()?;
     to_pydantic_dataclass(word, class)
 }
-pub fn to_rust_word_type(word_type: &PyAny) -> PyResult<UserDictWordType> {
+pub(crate) fn to_rust_word_type(word_type: &PyAny) -> PyResult<UserDictWordType> {
     let name = word_type.getattr("name")?.extract::<String>()?;
 
     serde_json::from_value::<UserDictWordType>(json!(name)).into_py_value_result()
 }
 
 #[ext(VoicevoxCoreResultExt)]
-pub impl<T> voicevox_core::Result<T> {
+pub(crate) impl<T> voicevox_core::Result<T> {
     fn into_py_result(self, py: Python<'_>) -> PyResult<T> {
         use voicevox_core::ErrorKind;
 
diff --git a/crates/voicevox_core_python_api/src/lib.rs b/crates/voicevox_core_python_api/src/lib.rs
index 4cde8d711..4d190333d 100644
--- a/crates/voicevox_core_python_api/src/lib.rs
+++ b/crates/voicevox_core_python_api/src/lib.rs
@@ -1,25 +1,15 @@
-use std::{marker::PhantomData, sync::Arc};
+use std::marker::PhantomData;
 
 mod convert;
-use self::convert::{
-    async_modify_accent_phrases, from_acceleration_mode, from_dataclass, from_utf8_path,
-    to_py_user_dict_word, to_py_uuid, to_pydantic_dataclass, to_pydantic_voice_model_meta,
-    to_rust_user_dict_word, to_rust_uuid, VoicevoxCoreResultExt as _,
-};
-use camino::Utf8PathBuf;
+use self::convert::{from_utf8_path, to_pydantic_dataclass, VoicevoxCoreResultExt as _};
 use easy_ext::ext;
 use log::debug;
 use pyo3::{
     create_exception,
     exceptions::{PyException, PyKeyError, PyValueError},
-    pyclass, pyfunction, pymethods, pymodule,
-    types::{IntoPyDict as _, PyBytes, PyDict, PyList, PyModule},
-    wrap_pyfunction, PyAny, PyObject, PyRef, PyResult, PyTypeInfo, Python, ToPyObject,
-};
-use uuid::Uuid;
-use voicevox_core::{
-    AccelerationMode, AudioQueryModel, InitializeOptions, StyleId, SynthesisOptions, TtsOptions,
-    UserDictWord, VoiceModelId,
+    pyfunction, pymodule,
+    types::PyModule,
+    wrap_pyfunction, PyAny, PyResult, PyTypeInfo, Python,
 };
 
 #[pymodule]
@@ -42,10 +32,10 @@ fn rust(py: Python<'_>, module: &PyModule) -> PyResult<()> {
     module.add_and_register_submodule(blocking_module)?;
 
     let asyncio_module = PyModule::new(py, "voicevox_core._rust.asyncio")?;
-    asyncio_module.add_class::<self::Synthesizer>()?;
-    asyncio_module.add_class::<self::OpenJtalk>()?;
-    asyncio_module.add_class::<self::VoiceModel>()?;
-    asyncio_module.add_class::<self::UserDict>()?;
+    asyncio_module.add_class::<self::asyncio::Synthesizer>()?;
+    asyncio_module.add_class::<self::asyncio::OpenJtalk>()?;
+    asyncio_module.add_class::<self::asyncio::VoiceModel>()?;
+    asyncio_module.add_class::<self::asyncio::UserDict>()?;
     module.add_and_register_submodule(asyncio_module)
 }
 
@@ -95,12 +85,6 @@ exceptions! {
     InvalidWordError: PyValueError;
 }
 
-#[pyclass]
-#[derive(Clone)]
-struct VoiceModel {
-    model: voicevox_core::tokio::VoiceModel,
-}
-
 #[pyfunction]
 fn supported_devices(py: Python<'_>) -> PyResult<&PyAny> {
     let class = py
@@ -111,395 +95,6 @@ fn supported_devices(py: Python<'_>) -> PyResult<&PyAny> {
     to_pydantic_dataclass(s, class)
 }
 
-#[pymethods]
-impl VoiceModel {
-    #[staticmethod]
-    fn from_path(
-        py: Python<'_>,
-        #[pyo3(from_py_with = "from_utf8_path")] path: Utf8PathBuf,
-    ) -> PyResult<&PyAny> {
-        pyo3_asyncio::tokio::future_into_py(py, async move {
-            let model = voicevox_core::tokio::VoiceModel::from_path(path).await;
-            let model = Python::with_gil(|py| model.into_py_result(py))?;
-            Ok(Self { model })
-        })
-    }
-
-    #[getter]
-    fn id(&self) -> &str {
-        self.model.id().raw_voice_model_id()
-    }
-
-    #[getter]
-    fn metas<'py>(&self, py: Python<'py>) -> Vec<&'py PyAny> {
-        to_pydantic_voice_model_meta(self.model.metas(), py).unwrap()
-    }
-}
-
-#[pyclass]
-#[derive(Clone)]
-struct OpenJtalk {
-    open_jtalk: voicevox_core::tokio::OpenJtalk,
-}
-
-#[pymethods]
-impl OpenJtalk {
-    #[allow(clippy::new_ret_no_self)]
-    #[staticmethod]
-    fn new(
-        #[pyo3(from_py_with = "from_utf8_path")] open_jtalk_dict_dir: Utf8PathBuf,
-        py: Python<'_>,
-    ) -> PyResult<&PyAny> {
-        pyo3_asyncio::tokio::future_into_py(py, async move {
-            let open_jtalk = voicevox_core::tokio::OpenJtalk::new(open_jtalk_dict_dir).await;
-            let open_jtalk = Python::with_gil(|py| open_jtalk.into_py_result(py))?;
-            Ok(Self { open_jtalk })
-        })
-    }
-
-    fn use_user_dict<'py>(&self, user_dict: UserDict, py: Python<'py>) -> PyResult<&'py PyAny> {
-        let this = self.open_jtalk.clone();
-
-        pyo3_asyncio::tokio::future_into_py(py, async move {
-            let result = this.use_user_dict(&user_dict.dict).await;
-            Python::with_gil(|py| result.into_py_result(py))
-        })
-    }
-}
-
-#[pyclass]
-struct Synthesizer {
-    synthesizer: Closable<voicevox_core::tokio::Synthesizer<voicevox_core::tokio::OpenJtalk>, Self>,
-}
-
-#[pymethods]
-impl Synthesizer {
-    #[new]
-    #[pyo3(signature =(
-        open_jtalk,
-        acceleration_mode = InitializeOptions::default().acceleration_mode,
-        cpu_num_threads = InitializeOptions::default().cpu_num_threads,
-    ))]
-    fn new(
-        open_jtalk: OpenJtalk,
-        #[pyo3(from_py_with = "from_acceleration_mode")] acceleration_mode: AccelerationMode,
-        cpu_num_threads: u16,
-    ) -> PyResult<Self> {
-        let synthesizer = voicevox_core::tokio::Synthesizer::new(
-            open_jtalk.open_jtalk.clone(),
-            &InitializeOptions {
-                acceleration_mode,
-                cpu_num_threads,
-            },
-        );
-        let synthesizer = Python::with_gil(|py| synthesizer.into_py_result(py))?;
-        let synthesizer = Closable::new(synthesizer);
-        Ok(Self { synthesizer })
-    }
-
-    fn __repr__(&self) -> &'static str {
-        "Synthesizer { .. }"
-    }
-
-    fn __enter__(slf: PyRef<'_, Self>) -> PyResult<PyRef<'_, Self>> {
-        slf.synthesizer.get()?;
-        Ok(slf)
-    }
-
-    fn __exit__(
-        &mut self,
-        #[allow(unused_variables)] exc_type: &PyAny,
-        #[allow(unused_variables)] exc_value: &PyAny,
-        #[allow(unused_variables)] traceback: &PyAny,
-    ) {
-        self.close();
-    }
-
-    #[getter]
-    fn is_gpu_mode(&self) -> PyResult<bool> {
-        let synthesizer = self.synthesizer.get()?;
-        Ok(synthesizer.is_gpu_mode())
-    }
-
-    #[getter]
-    fn metas<'py>(&self, py: Python<'py>) -> PyResult<Vec<&'py PyAny>> {
-        let synthesizer = self.synthesizer.get()?;
-        to_pydantic_voice_model_meta(&synthesizer.metas(), py)
-    }
-
-    fn load_voice_model<'py>(
-        &mut self,
-        model: &'py PyAny,
-        py: Python<'py>,
-    ) -> PyResult<&'py PyAny> {
-        let model: VoiceModel = model.extract()?;
-        let synthesizer = self.synthesizer.get()?.clone();
-        pyo3_asyncio::tokio::future_into_py(py, async move {
-            let result = synthesizer.load_voice_model(&model.model).await;
-            Python::with_gil(|py| result.into_py_result(py))
-        })
-    }
-
-    fn unload_voice_model(&mut self, voice_model_id: &str, py: Python<'_>) -> PyResult<()> {
-        self.synthesizer
-            .get()?
-            .unload_voice_model(&VoiceModelId::new(voice_model_id.to_string()))
-            .into_py_result(py)
-    }
-
-    fn is_loaded_voice_model(&self, voice_model_id: &str) -> PyResult<bool> {
-        Ok(self
-            .synthesizer
-            .get()?
-            .is_loaded_voice_model(&VoiceModelId::new(voice_model_id.to_string())))
-    }
-
-    fn audio_query_from_kana<'py>(
-        &self,
-        kana: &str,
-        style_id: u32,
-        py: Python<'py>,
-    ) -> PyResult<&'py PyAny> {
-        let synthesizer = self.synthesizer.get()?.clone();
-        let kana = kana.to_owned();
-        pyo3_asyncio::tokio::future_into_py_with_locals(
-            py,
-            pyo3_asyncio::tokio::get_current_locals(py)?,
-            async move {
-                let audio_query = synthesizer
-                    .audio_query_from_kana(&kana, StyleId::new(style_id))
-                    .await;
-
-                Python::with_gil(|py| {
-                    let class = py.import("voicevox_core")?.getattr("AudioQuery")?;
-                    let ret = to_pydantic_dataclass(audio_query.into_py_result(py)?, class)?;
-                    Ok(ret.to_object(py))
-                })
-            },
-        )
-    }
-
-    fn audio_query<'py>(&self, text: &str, style_id: u32, py: Python<'py>) -> PyResult<&'py PyAny> {
-        let synthesizer = self.synthesizer.get()?.clone();
-        let text = text.to_owned();
-        pyo3_asyncio::tokio::future_into_py_with_locals(
-            py,
-            pyo3_asyncio::tokio::get_current_locals(py)?,
-            async move {
-                let audio_query = synthesizer.audio_query(&text, StyleId::new(style_id)).await;
-
-                Python::with_gil(|py| {
-                    let audio_query = audio_query.into_py_result(py)?;
-                    let class = py.import("voicevox_core")?.getattr("AudioQuery")?;
-                    let ret = to_pydantic_dataclass(audio_query, class)?;
-                    Ok(ret.to_object(py))
-                })
-            },
-        )
-    }
-
-    fn create_accent_phrases_from_kana<'py>(
-        &self,
-        kana: &str,
-        style_id: u32,
-        py: Python<'py>,
-    ) -> PyResult<&'py PyAny> {
-        let synthesizer = self.synthesizer.get()?.clone();
-        let kana = kana.to_owned();
-        pyo3_asyncio::tokio::future_into_py_with_locals(
-            py,
-            pyo3_asyncio::tokio::get_current_locals(py)?,
-            async move {
-                let accent_phrases = synthesizer
-                    .create_accent_phrases_from_kana(&kana, StyleId::new(style_id))
-                    .await;
-                Python::with_gil(|py| {
-                    let class = py.import("voicevox_core")?.getattr("AccentPhrase")?;
-                    let accent_phrases = accent_phrases
-                        .into_py_result(py)?
-                        .iter()
-                        .map(|ap| to_pydantic_dataclass(ap, class))
-                        .collect::<PyResult<Vec<_>>>();
-                    let list = PyList::new(py, accent_phrases);
-                    Ok(list.to_object(py))
-                })
-            },
-        )
-    }
-
-    fn create_accent_phrases<'py>(
-        &self,
-        text: &str,
-        style_id: u32,
-        py: Python<'py>,
-    ) -> PyResult<&'py PyAny> {
-        let synthesizer = self.synthesizer.get()?.clone();
-        let text = text.to_owned();
-        pyo3_asyncio::tokio::future_into_py_with_locals(
-            py,
-            pyo3_asyncio::tokio::get_current_locals(py)?,
-            async move {
-                let accent_phrases = synthesizer
-                    .create_accent_phrases(&text, StyleId::new(style_id))
-                    .await;
-                Python::with_gil(|py| {
-                    let class = py.import("voicevox_core")?.getattr("AccentPhrase")?;
-                    let accent_phrases = accent_phrases
-                        .into_py_result(py)?
-                        .iter()
-                        .map(|ap| to_pydantic_dataclass(ap, class))
-                        .collect::<PyResult<Vec<_>>>();
-                    let list = PyList::new(py, accent_phrases);
-                    Ok(list.to_object(py))
-                })
-            },
-        )
-    }
-
-    fn replace_mora_data<'py>(
-        &self,
-        accent_phrases: &'py PyList,
-        style_id: u32,
-        py: Python<'py>,
-    ) -> PyResult<&'py PyAny> {
-        let synthesizer = self.synthesizer.get()?.clone();
-        async_modify_accent_phrases(
-            accent_phrases,
-            StyleId::new(style_id),
-            py,
-            |a, s| async move { synthesizer.replace_mora_data(&a, s).await },
-        )
-    }
-
-    fn replace_phoneme_length<'py>(
-        &self,
-        accent_phrases: &'py PyList,
-        style_id: u32,
-        py: Python<'py>,
-    ) -> PyResult<&'py PyAny> {
-        let synthesizer = self.synthesizer.get()?.clone();
-        async_modify_accent_phrases(
-            accent_phrases,
-            StyleId::new(style_id),
-            py,
-            |a, s| async move { synthesizer.replace_phoneme_length(&a, s).await },
-        )
-    }
-
-    fn replace_mora_pitch<'py>(
-        &self,
-        accent_phrases: &'py PyList,
-        style_id: u32,
-        py: Python<'py>,
-    ) -> PyResult<&'py PyAny> {
-        let synthesizer = self.synthesizer.get()?.clone();
-        async_modify_accent_phrases(
-            accent_phrases,
-            StyleId::new(style_id),
-            py,
-            |a, s| async move { synthesizer.replace_mora_pitch(&a, s).await },
-        )
-    }
-
-    #[pyo3(signature=(audio_query,style_id,enable_interrogative_upspeak = TtsOptions::default().enable_interrogative_upspeak))]
-    fn synthesis<'py>(
-        &self,
-        #[pyo3(from_py_with = "from_dataclass")] audio_query: AudioQueryModel,
-        style_id: u32,
-        enable_interrogative_upspeak: bool,
-        py: Python<'py>,
-    ) -> PyResult<&'py PyAny> {
-        let synthesizer = self.synthesizer.get()?.clone();
-        pyo3_asyncio::tokio::future_into_py_with_locals(
-            py,
-            pyo3_asyncio::tokio::get_current_locals(py)?,
-            async move {
-                let wav = synthesizer
-                    .synthesis(
-                        &audio_query,
-                        StyleId::new(style_id),
-                        &SynthesisOptions {
-                            enable_interrogative_upspeak,
-                        },
-                    )
-                    .await;
-                Python::with_gil(|py| {
-                    let wav = wav.into_py_result(py)?;
-                    Ok(PyBytes::new(py, &wav).to_object(py))
-                })
-            },
-        )
-    }
-
-    #[pyo3(signature=(
-        kana,
-        style_id,
-        enable_interrogative_upspeak = TtsOptions::default().enable_interrogative_upspeak
-    ))]
-    fn tts_from_kana<'py>(
-        &self,
-        kana: &str,
-        style_id: u32,
-        enable_interrogative_upspeak: bool,
-        py: Python<'py>,
-    ) -> PyResult<&'py PyAny> {
-        let style_id = StyleId::new(style_id);
-        let options = TtsOptions {
-            enable_interrogative_upspeak,
-        };
-        let synthesizer = self.synthesizer.get()?.clone();
-        let kana = kana.to_owned();
-        pyo3_asyncio::tokio::future_into_py_with_locals(
-            py,
-            pyo3_asyncio::tokio::get_current_locals(py)?,
-            async move {
-                let wav = synthesizer.tts_from_kana(&kana, style_id, &options).await;
-
-                Python::with_gil(|py| {
-                    let wav = wav.into_py_result(py)?;
-                    Ok(PyBytes::new(py, &wav).to_object(py))
-                })
-            },
-        )
-    }
-
-    #[pyo3(signature=(
-        text,
-        style_id,
-        enable_interrogative_upspeak = TtsOptions::default().enable_interrogative_upspeak
-    ))]
-    fn tts<'py>(
-        &self,
-        text: &str,
-        style_id: u32,
-        enable_interrogative_upspeak: bool,
-        py: Python<'py>,
-    ) -> PyResult<&'py PyAny> {
-        let style_id = StyleId::new(style_id);
-        let options = TtsOptions {
-            enable_interrogative_upspeak,
-        };
-        let synthesizer = self.synthesizer.get()?.clone();
-        let text = text.to_owned();
-        pyo3_asyncio::tokio::future_into_py_with_locals(
-            py,
-            pyo3_asyncio::tokio::get_current_locals(py)?,
-            async move {
-                let wav = synthesizer.tts(&text, style_id, &options).await;
-
-                Python::with_gil(|py| {
-                    let wav = wav.into_py_result(py)?;
-                    Ok(PyBytes::new(py, &wav).to_object(py))
-                })
-            },
-        )
-    }
-
-    fn close(&mut self) {
-        self.synthesizer.close()
-    }
-}
-
 struct Closable<T, C: PyTypeInfo> {
     content: MaybeClosed<T>,
     marker: PhantomData<C>,
@@ -552,96 +147,13 @@ fn _to_zenkaku(text: &str) -> PyResult<String> {
     Ok(voicevox_core::__internal::to_zenkaku(text))
 }
 
-#[pyclass]
-#[derive(Default, Debug, Clone)]
-struct UserDict {
-    dict: Arc<voicevox_core::tokio::UserDict>,
-}
-
-#[pymethods]
-impl UserDict {
-    #[new]
-    fn new() -> Self {
-        Self::default()
-    }
-
-    fn load<'py>(&self, path: &str, py: Python<'py>) -> PyResult<&'py PyAny> {
-        let this = self.dict.clone();
-        let path = path.to_owned();
-
-        pyo3_asyncio::tokio::future_into_py(py, async move {
-            let result = this.load(&path).await;
-            Python::with_gil(|py| result.into_py_result(py))
-        })
-    }
-
-    fn save<'py>(&self, path: &str, py: Python<'py>) -> PyResult<&'py PyAny> {
-        let this = self.dict.clone();
-        let path = path.to_owned();
-
-        pyo3_asyncio::tokio::future_into_py(py, async move {
-            let result = this.save(&path).await;
-            Python::with_gil(|py| result.into_py_result(py))
-        })
-    }
-
-    fn add_word(
-        &mut self,
-        #[pyo3(from_py_with = "to_rust_user_dict_word")] word: UserDictWord,
-        py: Python<'_>,
-    ) -> PyResult<PyObject> {
-        let uuid = self.dict.add_word(word).into_py_result(py)?;
-
-        to_py_uuid(py, uuid)
-    }
-
-    fn update_word(
-        &mut self,
-        #[pyo3(from_py_with = "to_rust_uuid")] word_uuid: Uuid,
-        #[pyo3(from_py_with = "to_rust_user_dict_word")] word: UserDictWord,
-        py: Python<'_>,
-    ) -> PyResult<()> {
-        self.dict.update_word(word_uuid, word).into_py_result(py)?;
-        Ok(())
-    }
-
-    fn remove_word(
-        &mut self,
-        #[pyo3(from_py_with = "to_rust_uuid")] word_uuid: Uuid,
-        py: Python<'_>,
-    ) -> PyResult<()> {
-        self.dict.remove_word(word_uuid).into_py_result(py)?;
-        Ok(())
-    }
-
-    fn import_dict(&mut self, other: &UserDict, py: Python<'_>) -> PyResult<()> {
-        self.dict.import(&other.dict).into_py_result(py)?;
-        Ok(())
-    }
-
-    #[getter]
-    fn words<'py>(&self, py: Python<'py>) -> PyResult<&'py PyDict> {
-        let words = self.dict.with_words(|words| {
-            words
-                .iter()
-                .map(|(&uuid, word)| {
-                    let uuid = to_py_uuid(py, uuid)?;
-                    let word = to_py_user_dict_word(py, word)?;
-                    Ok((uuid, word))
-                })
-                .collect::<PyResult<Vec<_>>>()
-        })?;
-        Ok(words.into_py_dict(py))
-    }
-}
-
 mod blocking {
-    use std::sync::Arc;
+    use std::{path::PathBuf, sync::Arc};
 
     use camino::Utf8PathBuf;
     use pyo3::{
         pyclass, pymethods,
-        types::{IntoPyDict as _, PyBytes, PyDict, PyList},
+        types::{IntoPyDict as _, PyBytes, PyDict, PyList, PyString},
         PyAny, PyObject, PyRef, PyResult, Python,
     };
     use uuid::Uuid;
@@ -661,10 +173,7 @@ mod blocking {
     #[pymethods]
     impl VoiceModel {
         #[staticmethod]
-        fn from_path(
-            py: Python<'_>,
-            #[pyo3(from_py_with = "crate::convert::from_utf8_path")] path: Utf8PathBuf,
-        ) -> PyResult<Self> {
+        fn from_path(py: Python<'_>, path: PathBuf) -> PyResult<Self> {
             let model = voicevox_core::blocking::VoiceModel::from_path(path).into_py_result(py)?;
             Ok(Self { model })
         }
@@ -786,7 +295,12 @@ mod blocking {
                 .into_py_result(py)
         }
 
-        fn is_loaded_voice_model(&self, voice_model_id: &str) -> PyResult<bool> {
+        // C APIの挙動と一貫性を持たせる。
+        fn is_loaded_voice_model(&self, voice_model_id: &PyString) -> PyResult<bool> {
+            let Ok(voice_model_id) = voice_model_id.to_str() else {
+                // 与えられたIDがUTF-8ではない場合、それに対応する`VoicdModel`は確実に存在しない
+                return Ok(false);
+            };
             Ok(self
                 .synthesizer
                 .get()?
@@ -1057,3 +571,512 @@ mod blocking {
         }
     }
 }
+
+mod asyncio {
+    use std::{path::PathBuf, sync::Arc};
+
+    use camino::Utf8PathBuf;
+    use pyo3::{
+        pyclass, pymethods,
+        types::{IntoPyDict as _, PyBytes, PyDict, PyList, PyString},
+        PyAny, PyObject, PyRef, PyResult, Python, ToPyObject as _,
+    };
+    use uuid::Uuid;
+    use voicevox_core::{
+        AccelerationMode, AudioQueryModel, InitializeOptions, StyleId, SynthesisOptions,
+        TtsOptions, UserDictWord, VoiceModelId,
+    };
+
+    use crate::{convert::VoicevoxCoreResultExt as _, Closable};
+
+    #[pyclass]
+    #[derive(Clone)]
+    pub(crate) struct VoiceModel {
+        model: voicevox_core::tokio::VoiceModel,
+    }
+
+    #[pymethods]
+    impl VoiceModel {
+        #[staticmethod]
+        fn from_path(py: Python<'_>, path: PathBuf) -> PyResult<&PyAny> {
+            pyo3_asyncio::tokio::future_into_py(py, async move {
+                let model = voicevox_core::tokio::VoiceModel::from_path(path).await;
+                let model = Python::with_gil(|py| model.into_py_result(py))?;
+                Ok(Self { model })
+            })
+        }
+
+        #[getter]
+        fn id(&self) -> &str {
+            self.model.id().raw_voice_model_id()
+        }
+
+        #[getter]
+        fn metas<'py>(&self, py: Python<'py>) -> Vec<&'py PyAny> {
+            crate::convert::to_pydantic_voice_model_meta(self.model.metas(), py).unwrap()
+        }
+    }
+
+    #[pyclass]
+    #[derive(Clone)]
+    pub(crate) struct OpenJtalk {
+        open_jtalk: voicevox_core::tokio::OpenJtalk,
+    }
+
+    #[pymethods]
+    impl OpenJtalk {
+        #[allow(clippy::new_ret_no_self)]
+        #[staticmethod]
+        fn new(
+            #[pyo3(from_py_with = "crate::convert::from_utf8_path")]
+            open_jtalk_dict_dir: Utf8PathBuf,
+            py: Python<'_>,
+        ) -> PyResult<&PyAny> {
+            pyo3_asyncio::tokio::future_into_py(py, async move {
+                let open_jtalk = voicevox_core::tokio::OpenJtalk::new(open_jtalk_dict_dir).await;
+                let open_jtalk = Python::with_gil(|py| open_jtalk.into_py_result(py))?;
+                Ok(Self { open_jtalk })
+            })
+        }
+
+        fn use_user_dict<'py>(&self, user_dict: UserDict, py: Python<'py>) -> PyResult<&'py PyAny> {
+            let this = self.open_jtalk.clone();
+
+            pyo3_asyncio::tokio::future_into_py(py, async move {
+                let result = this.use_user_dict(&user_dict.dict).await;
+                Python::with_gil(|py| result.into_py_result(py))
+            })
+        }
+    }
+
+    #[pyclass]
+    pub(crate) struct Synthesizer {
+        synthesizer:
+            Closable<voicevox_core::tokio::Synthesizer<voicevox_core::tokio::OpenJtalk>, Self>,
+    }
+
+    #[pymethods]
+    impl Synthesizer {
+        #[new]
+        #[pyo3(signature =(
+            open_jtalk,
+            acceleration_mode = InitializeOptions::default().acceleration_mode,
+            cpu_num_threads = InitializeOptions::default().cpu_num_threads,
+        ))]
+        fn new(
+            open_jtalk: OpenJtalk,
+            #[pyo3(from_py_with = "crate::convert::from_acceleration_mode")]
+            acceleration_mode: AccelerationMode,
+            cpu_num_threads: u16,
+        ) -> PyResult<Self> {
+            let synthesizer = voicevox_core::tokio::Synthesizer::new(
+                open_jtalk.open_jtalk.clone(),
+                &InitializeOptions {
+                    acceleration_mode,
+                    cpu_num_threads,
+                },
+            );
+            let synthesizer = Python::with_gil(|py| synthesizer.into_py_result(py))?;
+            let synthesizer = Closable::new(synthesizer);
+            Ok(Self { synthesizer })
+        }
+
+        fn __repr__(&self) -> &'static str {
+            "Synthesizer { .. }"
+        }
+
+        fn __enter__(slf: PyRef<'_, Self>) -> PyResult<PyRef<'_, Self>> {
+            slf.synthesizer.get()?;
+            Ok(slf)
+        }
+
+        fn __exit__(
+            &mut self,
+            #[allow(unused_variables)] exc_type: &PyAny,
+            #[allow(unused_variables)] exc_value: &PyAny,
+            #[allow(unused_variables)] traceback: &PyAny,
+        ) {
+            self.close();
+        }
+
+        #[getter]
+        fn is_gpu_mode(&self) -> PyResult<bool> {
+            let synthesizer = self.synthesizer.get()?;
+            Ok(synthesizer.is_gpu_mode())
+        }
+
+        #[getter]
+        fn metas<'py>(&self, py: Python<'py>) -> PyResult<Vec<&'py PyAny>> {
+            let synthesizer = self.synthesizer.get()?;
+            crate::convert::to_pydantic_voice_model_meta(&synthesizer.metas(), py)
+        }
+
+        fn load_voice_model<'py>(
+            &mut self,
+            model: &'py PyAny,
+            py: Python<'py>,
+        ) -> PyResult<&'py PyAny> {
+            let model: VoiceModel = model.extract()?;
+            let synthesizer = self.synthesizer.get()?.clone();
+            pyo3_asyncio::tokio::future_into_py(py, async move {
+                let result = synthesizer.load_voice_model(&model.model).await;
+                Python::with_gil(|py| result.into_py_result(py))
+            })
+        }
+
+        fn unload_voice_model(&mut self, voice_model_id: &str, py: Python<'_>) -> PyResult<()> {
+            self.synthesizer
+                .get()?
+                .unload_voice_model(&VoiceModelId::new(voice_model_id.to_string()))
+                .into_py_result(py)
+        }
+
+        // C APIの挙動と一貫性を持たせる。
+        fn is_loaded_voice_model(&self, voice_model_id: &PyString) -> PyResult<bool> {
+            let Ok(voice_model_id) = voice_model_id.to_str() else {
+                // 与えられたIDがUTF-8ではない場合、それに対応する`VoicdModel`は確実に存在しない
+                return Ok(false);
+            };
+            Ok(self
+                .synthesizer
+                .get()?
+                .is_loaded_voice_model(&VoiceModelId::new(voice_model_id.to_string())))
+        }
+
+        fn audio_query_from_kana<'py>(
+            &self,
+            kana: &str,
+            style_id: u32,
+            py: Python<'py>,
+        ) -> PyResult<&'py PyAny> {
+            let synthesizer = self.synthesizer.get()?.clone();
+            let kana = kana.to_owned();
+            pyo3_asyncio::tokio::future_into_py_with_locals(
+                py,
+                pyo3_asyncio::tokio::get_current_locals(py)?,
+                async move {
+                    let audio_query = synthesizer
+                        .audio_query_from_kana(&kana, StyleId::new(style_id))
+                        .await;
+
+                    Python::with_gil(|py| {
+                        let class = py.import("voicevox_core")?.getattr("AudioQuery")?;
+                        let ret = crate::convert::to_pydantic_dataclass(
+                            audio_query.into_py_result(py)?,
+                            class,
+                        )?;
+                        Ok(ret.to_object(py))
+                    })
+                },
+            )
+        }
+
+        fn audio_query<'py>(
+            &self,
+            text: &str,
+            style_id: u32,
+            py: Python<'py>,
+        ) -> PyResult<&'py PyAny> {
+            let synthesizer = self.synthesizer.get()?.clone();
+            let text = text.to_owned();
+            pyo3_asyncio::tokio::future_into_py_with_locals(
+                py,
+                pyo3_asyncio::tokio::get_current_locals(py)?,
+                async move {
+                    let audio_query = synthesizer.audio_query(&text, StyleId::new(style_id)).await;
+
+                    Python::with_gil(|py| {
+                        let audio_query = audio_query.into_py_result(py)?;
+                        let class = py.import("voicevox_core")?.getattr("AudioQuery")?;
+                        let ret = crate::convert::to_pydantic_dataclass(audio_query, class)?;
+                        Ok(ret.to_object(py))
+                    })
+                },
+            )
+        }
+
+        fn create_accent_phrases_from_kana<'py>(
+            &self,
+            kana: &str,
+            style_id: u32,
+            py: Python<'py>,
+        ) -> PyResult<&'py PyAny> {
+            let synthesizer = self.synthesizer.get()?.clone();
+            let kana = kana.to_owned();
+            pyo3_asyncio::tokio::future_into_py_with_locals(
+                py,
+                pyo3_asyncio::tokio::get_current_locals(py)?,
+                async move {
+                    let accent_phrases = synthesizer
+                        .create_accent_phrases_from_kana(&kana, StyleId::new(style_id))
+                        .await;
+                    Python::with_gil(|py| {
+                        let class = py.import("voicevox_core")?.getattr("AccentPhrase")?;
+                        let accent_phrases = accent_phrases
+                            .into_py_result(py)?
+                            .iter()
+                            .map(|ap| crate::convert::to_pydantic_dataclass(ap, class))
+                            .collect::<PyResult<Vec<_>>>();
+                        let list = PyList::new(py, accent_phrases);
+                        Ok(list.to_object(py))
+                    })
+                },
+            )
+        }
+
+        fn create_accent_phrases<'py>(
+            &self,
+            text: &str,
+            style_id: u32,
+            py: Python<'py>,
+        ) -> PyResult<&'py PyAny> {
+            let synthesizer = self.synthesizer.get()?.clone();
+            let text = text.to_owned();
+            pyo3_asyncio::tokio::future_into_py_with_locals(
+                py,
+                pyo3_asyncio::tokio::get_current_locals(py)?,
+                async move {
+                    let accent_phrases = synthesizer
+                        .create_accent_phrases(&text, StyleId::new(style_id))
+                        .await;
+                    Python::with_gil(|py| {
+                        let class = py.import("voicevox_core")?.getattr("AccentPhrase")?;
+                        let accent_phrases = accent_phrases
+                            .into_py_result(py)?
+                            .iter()
+                            .map(|ap| crate::convert::to_pydantic_dataclass(ap, class))
+                            .collect::<PyResult<Vec<_>>>();
+                        let list = PyList::new(py, accent_phrases);
+                        Ok(list.to_object(py))
+                    })
+                },
+            )
+        }
+
+        fn replace_mora_data<'py>(
+            &self,
+            accent_phrases: &'py PyList,
+            style_id: u32,
+            py: Python<'py>,
+        ) -> PyResult<&'py PyAny> {
+            let synthesizer = self.synthesizer.get()?.clone();
+            crate::convert::async_modify_accent_phrases(
+                accent_phrases,
+                StyleId::new(style_id),
+                py,
+                |a, s| async move { synthesizer.replace_mora_data(&a, s).await },
+            )
+        }
+
+        fn replace_phoneme_length<'py>(
+            &self,
+            accent_phrases: &'py PyList,
+            style_id: u32,
+            py: Python<'py>,
+        ) -> PyResult<&'py PyAny> {
+            let synthesizer = self.synthesizer.get()?.clone();
+            crate::convert::async_modify_accent_phrases(
+                accent_phrases,
+                StyleId::new(style_id),
+                py,
+                |a, s| async move { synthesizer.replace_phoneme_length(&a, s).await },
+            )
+        }
+
+        fn replace_mora_pitch<'py>(
+            &self,
+            accent_phrases: &'py PyList,
+            style_id: u32,
+            py: Python<'py>,
+        ) -> PyResult<&'py PyAny> {
+            let synthesizer = self.synthesizer.get()?.clone();
+            crate::convert::async_modify_accent_phrases(
+                accent_phrases,
+                StyleId::new(style_id),
+                py,
+                |a, s| async move { synthesizer.replace_mora_pitch(&a, s).await },
+            )
+        }
+
+        #[pyo3(signature=(audio_query,style_id,enable_interrogative_upspeak = TtsOptions::default().enable_interrogative_upspeak))]
+        fn synthesis<'py>(
+            &self,
+            #[pyo3(from_py_with = "crate::convert::from_dataclass")] audio_query: AudioQueryModel,
+            style_id: u32,
+            enable_interrogative_upspeak: bool,
+            py: Python<'py>,
+        ) -> PyResult<&'py PyAny> {
+            let synthesizer = self.synthesizer.get()?.clone();
+            pyo3_asyncio::tokio::future_into_py_with_locals(
+                py,
+                pyo3_asyncio::tokio::get_current_locals(py)?,
+                async move {
+                    let wav = synthesizer
+                        .synthesis(
+                            &audio_query,
+                            StyleId::new(style_id),
+                            &SynthesisOptions {
+                                enable_interrogative_upspeak,
+                            },
+                        )
+                        .await;
+                    Python::with_gil(|py| {
+                        let wav = wav.into_py_result(py)?;
+                        Ok(PyBytes::new(py, &wav).to_object(py))
+                    })
+                },
+            )
+        }
+
+        #[pyo3(signature=(
+            kana,
+            style_id,
+            enable_interrogative_upspeak = TtsOptions::default().enable_interrogative_upspeak
+        ))]
+        fn tts_from_kana<'py>(
+            &self,
+            kana: &str,
+            style_id: u32,
+            enable_interrogative_upspeak: bool,
+            py: Python<'py>,
+        ) -> PyResult<&'py PyAny> {
+            let style_id = StyleId::new(style_id);
+            let options = TtsOptions {
+                enable_interrogative_upspeak,
+            };
+            let synthesizer = self.synthesizer.get()?.clone();
+            let kana = kana.to_owned();
+            pyo3_asyncio::tokio::future_into_py_with_locals(
+                py,
+                pyo3_asyncio::tokio::get_current_locals(py)?,
+                async move {
+                    let wav = synthesizer.tts_from_kana(&kana, style_id, &options).await;
+
+                    Python::with_gil(|py| {
+                        let wav = wav.into_py_result(py)?;
+                        Ok(PyBytes::new(py, &wav).to_object(py))
+                    })
+                },
+            )
+        }
+
+        #[pyo3(signature=(
+            text,
+            style_id,
+            enable_interrogative_upspeak = TtsOptions::default().enable_interrogative_upspeak
+        ))]
+        fn tts<'py>(
+            &self,
+            text: &str,
+            style_id: u32,
+            enable_interrogative_upspeak: bool,
+            py: Python<'py>,
+        ) -> PyResult<&'py PyAny> {
+            let style_id = StyleId::new(style_id);
+            let options = TtsOptions {
+                enable_interrogative_upspeak,
+            };
+            let synthesizer = self.synthesizer.get()?.clone();
+            let text = text.to_owned();
+            pyo3_asyncio::tokio::future_into_py_with_locals(
+                py,
+                pyo3_asyncio::tokio::get_current_locals(py)?,
+                async move {
+                    let wav = synthesizer.tts(&text, style_id, &options).await;
+
+                    Python::with_gil(|py| {
+                        let wav = wav.into_py_result(py)?;
+                        Ok(PyBytes::new(py, &wav).to_object(py))
+                    })
+                },
+            )
+        }
+
+        fn close(&mut self) {
+            self.synthesizer.close()
+        }
+    }
+
+    #[pyclass]
+    #[derive(Default, Debug, Clone)]
+    pub(crate) struct UserDict {
+        dict: Arc<voicevox_core::tokio::UserDict>,
+    }
+
+    #[pymethods]
+    impl UserDict {
+        #[new]
+        fn new() -> Self {
+            Self::default()
+        }
+
+        fn load<'py>(&self, path: &str, py: Python<'py>) -> PyResult<&'py PyAny> {
+            let this = self.dict.clone();
+            let path = path.to_owned();
+
+            pyo3_asyncio::tokio::future_into_py(py, async move {
+                let result = this.load(&path).await;
+                Python::with_gil(|py| result.into_py_result(py))
+            })
+        }
+
+        fn save<'py>(&self, path: &str, py: Python<'py>) -> PyResult<&'py PyAny> {
+            let this = self.dict.clone();
+            let path = path.to_owned();
+
+            pyo3_asyncio::tokio::future_into_py(py, async move {
+                let result = this.save(&path).await;
+                Python::with_gil(|py| result.into_py_result(py))
+            })
+        }
+
+        fn add_word(
+            &mut self,
+            #[pyo3(from_py_with = "crate::convert::to_rust_user_dict_word")] word: UserDictWord,
+            py: Python<'_>,
+        ) -> PyResult<PyObject> {
+            let uuid = self.dict.add_word(word).into_py_result(py)?;
+
+            crate::convert::to_py_uuid(py, uuid)
+        }
+
+        fn update_word(
+            &mut self,
+            #[pyo3(from_py_with = "crate::convert::to_rust_uuid")] word_uuid: Uuid,
+            #[pyo3(from_py_with = "crate::convert::to_rust_user_dict_word")] word: UserDictWord,
+            py: Python<'_>,
+        ) -> PyResult<()> {
+            self.dict.update_word(word_uuid, word).into_py_result(py)?;
+            Ok(())
+        }
+
+        fn remove_word(
+            &mut self,
+            #[pyo3(from_py_with = "crate::convert::to_rust_uuid")] word_uuid: Uuid,
+            py: Python<'_>,
+        ) -> PyResult<()> {
+            self.dict.remove_word(word_uuid).into_py_result(py)?;
+            Ok(())
+        }
+
+        fn import_dict(&mut self, other: &UserDict, py: Python<'_>) -> PyResult<()> {
+            self.dict.import(&other.dict).into_py_result(py)?;
+            Ok(())
+        }
+
+        #[getter]
+        fn words<'py>(&self, py: Python<'py>) -> PyResult<&'py PyDict> {
+            let words = self.dict.with_words(|words| {
+                words
+                    .iter()
+                    .map(|(&uuid, word)| {
+                        let uuid = crate::convert::to_py_uuid(py, uuid)?;
+                        let word = crate::convert::to_py_user_dict_word(py, word)?;
+                        Ok((uuid, word))
+                    })
+                    .collect::<PyResult<Vec<_>>>()
+            })?;
+            Ok(words.into_py_dict(py))
+        }
+    }
+}
diff --git a/docs/vvm.md b/docs/vvm.md
index 572ffee1a..c2de7fb41 100644
--- a/docs/vvm.md
+++ b/docs/vvm.md
@@ -1,9 +1,24 @@
 # VVM ファイル
 
-音声合成するために必要な onnx モデルファイルなどがまとめられた zip 形式のファイル。
-root パスに確定で`manifest.json`を持つ。
+***VVM ファイル*** は、音声合成に必要な声情報を含むファイルである。  
+
+より正確には、音声合成のモデル重みファイルなどを含む zip 形式のアーカイブである。拡張子は `.vvm`。  
+以下の内部ディレクトリ構造を持つ：  
+
+- `{filename}.vvm`
+  - `manifest.json`
+  - `metas.json`
+  - <duration_model>
+  - <intonation_model>
+  - <decode_model>
+
+model は `.onnx` や `.bin` など様々ある。例えば `sample.vvm` は `predict_duration.onnx` / `predict_intonation.onnx` / `decode.onnx` を含む。  
+
+VOICEVOX OSS が提供する VVM には [`sample.vvm`](https://github.com/VOICEVOX/voicevox_core/tree/main/model) がある。  
+製品版 VOICEVOX で利用される VVM は [こちらのレポジトリ](https://github.com/VOICEVOX/voicevox_fat_resource/tree/main/core/model) で確認できる。  
 
 ## マニフェストファイル
 
-ファイルの構成や、onnx モデルなどを読み込む・利用するのに必要な情報を記述した json ファイル。
-root パスに`manifest.json`として配置する。
+VVM における ***マニフェストファイル*** は、VVM ファイルの構成や、onnx モデルなどを読み込む・利用するのに必要な情報を記述したファイルである。  
+json 形式で記述され、root パスに`manifest.json`として配置する。  
+[VOICEVOX CORE のソースコード](https://github.com/VOICEVOX/voicevox_core/blob/main/crates/voicevox_core/src/manifest.rs) 内で `Manifest` 構造体としてスキーマが定義されている。  
diff --git a/rust-toolchain b/rust-toolchain
index 32a6ce3c7..b1131583c 100644
--- a/rust-toolchain
+++ b/rust-toolchain
@@ -1 +1 @@
-1.76.0
+1.77.1