Phoneme conversion presets

README.md

@@ -47,8 +47,7 @@ If you are developing OSS projects related to singing voice synthesis, you may f
 
 ## Contributors
 
-[sdercolin](https://github.com/sdercolin), [ghosrt](https://github.com/ghosrt)
-, [shine5402](https://github.com/shine5402)
+[sdercolin](https://github.com/sdercolin), [ghosrt](https://github.com/ghosrt), [shine5402](https://github.com/shine5402), [adlez27](https://github.com/adlez27)
 
 ## Localization help
 

src/jsMain/kotlin/model/Format.kt

@@ -135,7 +135,7 @@ enum class Format(
             io.Vsq.generate(project, features)
         },
         possibleLyricsTypes = listOf(RomajiCv, KanaCv),
-        availableFeaturesForGeneration = listOf(ConvertPitch),
+        availableFeaturesForGeneration = listOf(ConvertPitch, ConvertPhonemes),
     ),
     VocaloidMid(
         "mid",

src/jsMain/kotlin/model/PhonemesMappingPreset.kt

@@ -0,0 +1,10 @@
+package model
+
+import process.phonemes.PhonemesMappingRequest
+
+class PhonemesMappingPreset(
+    val sourceFormats: List<Format>,
+    val targetFormats: List<Format>,
+    val name: String,
+    val phonemesMap: PhonemesMappingRequest,
+)

src/jsMain/kotlin/process/phonemes/PhonemesMapping.kt

@@ -1,8 +1,11 @@
 package process.phonemes
 
+import external.require
 import kotlinx.serialization.Serializable
 import kotlinx.serialization.Transient
+import model.Format
 import model.Note
+import model.PhonemesMappingPreset
 import model.Project
 import model.Track
 import process.validateNotes
@@ -28,12 +31,189 @@ data class PhonemesMappingRequest(
 
     companion object {
 
-        fun findPreset(name: String) = Presets.find { it.first == name }?.second
+        fun findPreset(name: String) = Presets.find { it.name == name }?.phonemesMap
 
         fun getPreset(name: String) = requireNotNull(findPreset(name))
 
-        val Presets: List<Pair<String, PhonemesMappingRequest>> by lazy {
-            listOf()
+        val Presets: List<PhonemesMappingPreset> by lazy {
+            listOf(
+                PhonemesMappingPreset(
+                    sourceFormats = listOf(Format.Svp),
+                    targetFormats = Format.vocaloidFormats,
+                    name = "Japanese (SynthV to Vocaloid)",
+                    phonemesMap = PhonemesMappingRequest(
+                        require("./texts/SynthV JA to Vocaloid JA.txt").default as String,
+                    ),
+                ),
+                PhonemesMappingPreset(
+                    sourceFormats = Format.vocaloidFormats,
+                    targetFormats = listOf(Format.Svp),
+                    name = "Japanese (Vocaloid to SynthV)",
+                    phonemesMap = PhonemesMappingRequest(
+                        require("./texts/Vocaloid JA to SynthV JA.txt").default as String,
+                    ),
+                ),
+                PhonemesMappingPreset(
+                    sourceFormats = listOf(Format.Svp),
+                    targetFormats = Format.vocaloidFormats,
+                    name = "English (SynthV to Vocaloid)",
+                    phonemesMap = PhonemesMappingRequest(
+                        require("./texts/SynthV EN to Vocaloid EN.txt").default as String,
+                    ),
+                ),
+                PhonemesMappingPreset(
+                    sourceFormats = listOf(Format.Svp),
+                    targetFormats = listOf(Format.Ustx),
+                    name = "English: ARPAsing (SynthV to OpenUtau)",
+                    phonemesMap = PhonemesMappingRequest(
+                        require("./texts/SynthV EN to OpenUtau ARPAsing.txt").default as String,
+                    ),
+                ),
+                PhonemesMappingPreset(
+                    sourceFormats = listOf(Format.Svp),
+                    targetFormats = listOf(Format.Ustx),
+                    name = "English: X-SAMPA (SynthV to OpenUtau)",
+                    phonemesMap = PhonemesMappingRequest(
+                        require("./texts/SynthV EN to OpenUtau X-SAMPA EN.txt").default as String,
+                    ),
+                ),
+                PhonemesMappingPreset(
+                    sourceFormats = listOf(Format.Svp),
+                    targetFormats = listOf(Format.Ustx),
+                    name = "English: VCCV (SynthV to OpenUtau)",
+                    phonemesMap = PhonemesMappingRequest(
+                        require("./texts/SynthV EN to OpenUtau VCCV.txt").default as String,
+                    ),
+                ),
+                PhonemesMappingPreset(
+                    sourceFormats = Format.vocaloidFormats,
+                    targetFormats = listOf(Format.Svp),
+                    name = "English (Vocaloid to SynthV)",
+                    phonemesMap = PhonemesMappingRequest(
+                        require("./texts/Vocaloid EN to SynthV EN.txt").default as String,
+                    ),
+                ),
+                PhonemesMappingPreset(
+                    sourceFormats = Format.vocaloidFormats,
+                    targetFormats = listOf(Format.Ustx),
+                    name = "English: ARPAsing (Vocaloid to OpenUtau)",
+                    phonemesMap = PhonemesMappingRequest(
+                        require("./texts/Vocaloid EN to OpenUtau ARPAsing.txt").default as String,
+                    ),
+                ),
+                PhonemesMappingPreset(
+                    sourceFormats = Format.vocaloidFormats,
+                    targetFormats = listOf(Format.Ustx),
+                    name = "English: X-SAMPA (Vocaloid to OpenUtau)",
+                    phonemesMap = PhonemesMappingRequest(
+                        require("./texts/Vocaloid EN to OpenUtau X-SAMPA EN.txt").default as String,
+                    ),
+                ),
+                PhonemesMappingPreset(
+                    sourceFormats = Format.vocaloidFormats,
+                    targetFormats = listOf(Format.Ustx),
+                    name = "English: VCCV (Vocaloid to OpenUtau)",
+                    phonemesMap = PhonemesMappingRequest(
+                        require("./texts/Vocaloid EN to OpenUtau VCCV.txt").default as String,
+                    ),
+                ),
+                PhonemesMappingPreset(
+                    sourceFormats = listOf(Format.Ustx),
+                    targetFormats = listOf(Format.Svp),
+                    name = "English: ARPAsing (OpenUtau to SynthV)",
+                    phonemesMap = PhonemesMappingRequest(
+                        require("./texts/OpenUtau ARPAsing to SynthV EN.txt").default as String,
+                    ),
+                ),
+                PhonemesMappingPreset(
+                    sourceFormats = listOf(Format.Ustx),
+                    targetFormats = Format.vocaloidFormats,
+                    name = "English: ARPAsing (OpenUtau to Vocaloid)",
+                    phonemesMap = PhonemesMappingRequest(
+                        require("./texts/OpenUtau ARPAsing to Vocaloid EN.txt").default as String,
+                    ),
+                ),
+                PhonemesMappingPreset(
+                    sourceFormats = listOf(Format.Ustx),
+                    targetFormats = listOf(Format.Ustx),
+                    name = "English: ARPAsing to X-SAMPA (OpenUtau to OpenUtau)",
+                    phonemesMap = PhonemesMappingRequest(
+                        require("./texts/OpenUtau ARPAsing to OpenUtau X-SAMPA EN.txt").default as String,
+                    ),
+                ),
+                PhonemesMappingPreset(
+                    sourceFormats = listOf(Format.Ustx),
+                    targetFormats = listOf(Format.Ustx),
+                    name = "English: ARPAsing to VCCV (OpenUtau to OpenUtau)",
+                    phonemesMap = PhonemesMappingRequest(
+                        require("./texts/OpenUtau ARPAsing to OpenUtau VCCV.txt").default as String,
+                    ),
+                ),
+                PhonemesMappingPreset(
+                    sourceFormats = listOf(Format.Ustx),
+                    targetFormats = listOf(Format.Svp),
+                    name = "English: X-SAMPA (OpenUtau to SynthV)",
+                    phonemesMap = PhonemesMappingRequest(
+                        require("./texts/OpenUtau X-SAMPA EN to SynthV EN.txt").default as String,
+                    ),
+                ),
+                PhonemesMappingPreset(
+                    sourceFormats = listOf(Format.Ustx),
+                    targetFormats = Format.vocaloidFormats,
+                    name = "English: X-SAMPA (OpenUtau to Vocaloid)",
+                    phonemesMap = PhonemesMappingRequest(
+                        require("./texts/OpenUtau X-SAMPA EN to Vocaloid EN.txt").default as String,
+                    ),
+                ),
+                PhonemesMappingPreset(
+                    sourceFormats = listOf(Format.Ustx),
+                    targetFormats = listOf(Format.Ustx),
+                    name = "English: X-SAMPA to ARPAsing (OpenUtau to OpenUtau)",
+                    phonemesMap = PhonemesMappingRequest(
+                        require("./texts/OpenUtau X-SAMPA EN to OpenUtau ARPAsing.txt").default as String,
+                    ),
+                ),
+                PhonemesMappingPreset(
+                    sourceFormats = listOf(Format.Ustx),
+                    targetFormats = listOf(Format.Ustx),
+                    name = "English: X-SAMPA to VCCV (OpenUtau to OpenUtau)",
+                    phonemesMap = PhonemesMappingRequest(
+                        require("./texts/OpenUtau X-SAMPA EN to OpenUtau VCCV.txt").default as String,
+                    ),
+                ),
+                PhonemesMappingPreset(
+                    sourceFormats = listOf(Format.Ustx),
+                    targetFormats = listOf(Format.Svp),
+                    name = "English: VCCV (OpenUtau to SynthV)",
+                    phonemesMap = PhonemesMappingRequest(
+                        require("./texts/OpenUtau VCCV to SynthV EN.txt").default as String,
+                    ),
+                ),
+                PhonemesMappingPreset(
+                    sourceFormats = listOf(Format.Ustx),
+                    targetFormats = Format.vocaloidFormats,
+                    name = "English: VCCV (OpenUtau to Vocaloid)",
+                    phonemesMap = PhonemesMappingRequest(
+                        require("./texts/OpenUtau VCCV to Vocaloid EN.txt").default as String,
+                    ),
+                ),
+                PhonemesMappingPreset(
+                    sourceFormats = listOf(Format.Ustx),
+                    targetFormats = listOf(Format.Ustx),
+                    name = "English: VCCV to ARPAsing (OpenUtau to OpenUtau)",
+                    phonemesMap = PhonemesMappingRequest(
+                        require("./texts/OpenUtau VCCV to OpenUtau ARPAsing.txt").default as String,
+                    ),
+                ),
+                PhonemesMappingPreset(
+                    sourceFormats = listOf(Format.Ustx),
+                    targetFormats = listOf(Format.Ustx),
+                    name = "English: VCCV to X-SAMPA (OpenUtau to OpenUtau)",
+                    phonemesMap = PhonemesMappingRequest(
+                        require("./texts/OpenUtau VCCV to OpenUtau X-SAMPA EN.txt").default as String,
+                    ),
+                ),
+            )
         }
     }
 }
@@ -48,20 +228,22 @@ fun Track.replacePhonemes(request: PhonemesMappingRequest?) = copy(
 
 fun Note.replacePhonemes(request: PhonemesMappingRequest?): Note {
     if (request == null) return copy(phoneme = null)
-    var output = phoneme?.split(" ") ?: return this
+    val rawInput = phoneme?.split(" ") ?: return this
+    val input = rawInput.map { it to true }.toMutableList() // boolean represents the phoneme is not converted yet
+    var output = mutableListOf<Pair<String, Int>>() // int represents index from input phoneme list
     for ((key, value) in request.map) {
         val keySplit = key.split(" ")
-        for (i in 0..(output.size - keySplit.size)) {
-            val subList = output.subList(i, i + keySplit.size)
-            if (subList == keySplit) {
-                output = output.subList(0, i) + value.split(" ") +
-                    output.subList(i + keySplit.size, output.size)
-                // once replaced, the count of output is changed, so we could not proceed the replacement.
-                // this means that multiple occurrences of the same phoneme set in one note get replaced only once.
-                break
+        for (i in 0..(input.size - keySplit.size)) {
+            val subList = input.subList(i, i + keySplit.size)
+            if (subList.map { it.first } == keySplit && subList.map { it.second }.all { it }) {
+                output += value to i
+                for (j in 0 until subList.size) {
+                    subList[j] = subList[j].first to false
+                }
             }
         }
     }
-    output = output.filter { it.isNotBlank() }
-    return copy(phoneme = output.joinToString(" "))
+    output += input.mapIndexedNotNull { index, pair -> if (pair.second) pair.first to index else null }
+    output = output.filter { it.first.isNotBlank() }.sortedBy { it.second }.toMutableList()
+    return copy(phoneme = output.joinToString(" ") { it.first })
 }

src/jsMain/kotlin/ui/ConfigurationEditor.kt

@@ -235,6 +235,8 @@ val ConfigurationEditor = scopedFC<ConfigurationEditorProps> { props, scope ->
         submitState = setLyricsMapping
     }
     if (phonemesConversion.isAvailable) PhonemesConversionBlock {
+        this.sourceFormat = props.projects.first().format
+        this.targetFormat = props.outputFormat
         initialState = phonemesConversion
         submitState = setPhonemesConversion
     }

src/jsMain/kotlin/ui/configuration/PhonemesConversion.kt

@@ -11,6 +11,7 @@ import csstype.em
 import csstype.px
 import emotion.react.css
 import kotlinx.js.jso
+import model.Format
 import mui.icons.material.HelpOutline
 import mui.material.BaseTextFieldProps
 import mui.material.Button
@@ -40,9 +41,12 @@ import ui.common.subFC
 import ui.strings.Strings
 import ui.strings.string
 
-external interface PhonemesConversionProps : SubProps<PhonemesConversionState>
+external interface PhonemesConversionProps : SubProps<PhonemesConversionState> {
+    var sourceFormat: Format
+    var targetFormat: Format
+}
 
-val PhonemesConversionBlock = subFC<PhonemesConversionProps, PhonemesConversionState> { _, state, editState ->
+val PhonemesConversionBlock = subFC<PhonemesConversionProps, PhonemesConversionState> { props, state, editState ->
     FormGroup {
         div {
             configurationSwitch(
@@ -53,10 +57,11 @@ val PhonemesConversionBlock = subFC<PhonemesConversionProps, PhonemesConversionS
         }
     }
 
-    if (state.isOn) buildPhonemesConversionDetail(state, editState)
+    if (state.isOn) buildPhonemesConversionDetail(props, state, editState)
 }
 
 private fun ChildrenBuilder.buildPhonemesConversionDetail(
+    props: PhonemesConversionProps,
     state: PhonemesConversionState,
     editState: (PhonemesConversionState.() -> PhonemesConversionState) -> Unit,
 ) {
@@ -138,12 +143,17 @@ private fun ChildrenBuilder.buildPhonemesConversionDetail(
                                         )
                                     }
                                 }
-                                PhonemesMappingRequest.Presets.forEach { preset ->
-                                    MenuItem {
-                                        value = preset.first
-                                        +(preset.first)
+                                PhonemesMappingRequest.Presets
+                                    .filter { preset ->
+                                        props.sourceFormat in preset.sourceFormats + Format.UfData &&
+                                            props.targetFormat in preset.targetFormats + Format.UfData
+                                    }
+                                    .forEach { preset ->
+                                        MenuItem {
+                                            value = preset.name
+                                            +(preset.name)
+                                        }
                                     }
-                                }
                             }
                         }
                         Button {

src/jsMain/resources/texts/OpenUtau ARPAsing to OpenUtau VCCV.txt

@@ -0,0 +1,28 @@
+aa=a
+ae=@
+ah=u
+ao r=0 r
+ao=9
+aw=8
+ay=I
+eh=e
+er=3
+ey=A
+hh iy=hh E
+hh y=hh y
+hh=h
+ih=i
+iy=E
+jh=j
+ow l=0 l
+ow=O
+oy=Q
+s k=sk
+s p=sp
+s t=st
+s m=sm
+s n=sn
+uh=6
+uw=o
+ae n=& n
+ae m=& m

src/jsMain/resources/texts/OpenUtau ARPAsing to OpenUtau X-SAMPA EN.txt

@@ -0,0 +1,24 @@
+aa=A
+ae={
+ah=V
+ao=O
+aw=aU
+ay=aI
+ch=tS
+dh=D
+eh=E
+er=3
+ey=eI
+hh=h
+ih=I
+iy=i
+jh=dZ
+ng=N
+ow=oU
+oy=OI
+sh=S
+th=T
+uh=U
+uw=u
+zh=Z
+y=j

src/jsMain/resources/texts/OpenUtau ARPAsing to SynthV EN.txt

@@ -0,0 +1,2 @@
+d r=dr
+t r=tr