Implement phonemes conversion

src/jsMain/kotlin/io/Svp.kt

@@ -108,6 +108,7 @@ object Svp {
             tickOn = tickOn,
             tickOff = tickOn + note.duration / TICK_RATE,
             lyric = note.lyrics.takeUnless { it.isNullOrBlank() } ?: DEFAULT_LYRIC,
+            phoneme = note.phonemes,
         )
     }
 
@@ -250,7 +251,7 @@ object Svp {
                         onset = it.tickOn * TICK_RATE,
                         duration = it.length * TICK_RATE,
                         lyrics = it.lyric,
-                        phonemes = "",
+                        phonemes = it.phoneme ?: "",
                         pitch = it.key,
                         attributes = Attributes(),
                     )

src/jsMain/kotlin/io/Ustx.kt

@@ -94,10 +94,20 @@ object Ustx {
             val track = trackMap[trackId] ?: continue
             val tickPrefix = voicePart.position
             val notes = voicePart.notes.map {
+                val rawLyrics = it.lyric
+                // some phonemizers use `lyrics [phonemes]` format
+                val regex = Regex("\\s\\[([^\\[\\]]*)\\]$")
+                val match = regex.find(rawLyrics)
+                val (lyric, phoneme) = if (match == null) {
+                    rawLyrics to null
+                } else {
+                    rawLyrics.take(match.range.first) to rawLyrics.substring(match.range).trim().trim('[', ']')
+                }
                 Note(
                     id = 0,
                     key = it.tone,
-                    lyric = it.lyric,
+                    lyric = lyric,
+                    phoneme = phoneme,
                     tickOn = it.position + tickPrefix,
                     tickOff = it.position + it.duration + tickPrefix,
                 )
@@ -265,12 +275,18 @@ object Ustx {
             if (index == 0) datum.copy(y = firstPitchPointValue) else datum.copy()
         }
         val pitch = template.pitch.copy(data = pitchPoints)
+        val lyric = buildString {
+            append(thisNote.lyric)
+            if (thisNote.phoneme?.isNotBlank() == true) {
+                append(" [${thisNote.phoneme}]")
+            }
+        }
         return Note(
             position = thisNote.tickOn,
             duration = thisNote.length,
             tone = thisNote.key,
             pitch = pitch,
-            lyric = thisNote.lyric,
+            lyric = lyric,
             vibrato = template.vibrato.copy(),
         )
     }

src/jsMain/kotlin/model/Feature.kt

@@ -13,6 +13,11 @@ enum class Feature(val isAvailable: (Project) -> Boolean) {
     ),
     SplitProject(
         isAvailable = { true },
+    ),
+    ConvertPhonemes(
+        isAvailable = { project ->
+            project.tracks.any { track -> track.notes.any { note -> note.phoneme != null } }
+        },
     )
 }
 

src/jsMain/kotlin/model/Format.kt

@@ -1,6 +1,7 @@
 package model
 
 import io.VsqLike
+import model.Feature.ConvertPhonemes
 import model.Feature.ConvertPitch
 import model.Feature.SplitProject
 import model.JapaneseLyricsType.KanaCv
@@ -31,7 +32,7 @@ enum class Format(
             io.Vsqx.generate(project, features)
         },
         possibleLyricsTypes = listOf(RomajiCv, KanaCv),
-        availableFeaturesForGeneration = listOf(ConvertPitch),
+        availableFeaturesForGeneration = listOf(ConvertPitch, ConvertPhonemes),
     ),
     Vpr(
         "vpr",
@@ -42,7 +43,7 @@ enum class Format(
             io.Vpr.generate(project, features)
         },
         possibleLyricsTypes = listOf(RomajiCv, KanaCv),
-        availableFeaturesForGeneration = listOf(ConvertPitch),
+        availableFeaturesForGeneration = listOf(ConvertPitch, ConvertPhonemes),
     ),
     Ust(
         "ust",
@@ -66,7 +67,7 @@ enum class Format(
             io.Ustx.generate(project, features)
         },
         possibleLyricsTypes = listOf(RomajiCv, RomajiVcv, KanaCv, KanaVcv),
-        availableFeaturesForGeneration = listOf(ConvertPitch),
+        availableFeaturesForGeneration = listOf(ConvertPitch, ConvertPhonemes),
     ),
     Ccs(
         "ccs",
@@ -88,7 +89,7 @@ enum class Format(
             io.Svp.generate(project, features)
         },
         possibleLyricsTypes = listOf(RomajiCv, KanaCv),
-        availableFeaturesForGeneration = listOf(ConvertPitch, SplitProject),
+        availableFeaturesForGeneration = listOf(ConvertPitch, SplitProject, ConvertPhonemes),
     ),
     S5p(
         "s5p",
@@ -181,7 +182,7 @@ enum class Format(
             io.UfData.generate(project, features)
         },
         possibleLyricsTypes = listOf(RomajiCv, RomajiVcv, KanaCv, KanaVcv),
-        availableFeaturesForGeneration = listOf(ConvertPitch),
+        availableFeaturesForGeneration = listOf(ConvertPitch, ConvertPhonemes),
     );
 
     private val allExtensions get() = listOf(extension) + otherExtensions
@@ -235,5 +236,8 @@ enum class Format(
                 StandardMid,
                 UfData,
             )
+
+        val vocaloidFormats: List<Format>
+            get() = listOf(Vsq, Vsqx, VocaloidMid, Vpr)
     }
 }

src/jsMain/kotlin/process/lyrics/LyricsMapping.kt

@@ -44,7 +44,7 @@ fun Project.mapLyrics(request: LyricsMappingRequest) = copy(
     tracks = tracks.map { it.replaceLyrics(request) },
 )
 
-fun Track.replaceLyrics(request: LyricsMappingRequest) = copy(
+private fun Track.replaceLyrics(request: LyricsMappingRequest) = copy(
     notes = notes.mapNotNull { note -> note.replaceLyrics(request).takeIf { it.lyric.isNotEmpty() } }
         .validateNotes(),
 )

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

@@ -33,17 +33,18 @@ data class PhonemesMappingRequest(
     }
 }
 
-fun Project.mapPhonemes(request: PhonemesMappingRequest) = copy(
+fun Project.mapPhonemes(request: PhonemesMappingRequest?) = copy(
     tracks = tracks.map { it.replacePhonemes(request) },
 )
 
-fun Track.replacePhonemes(request: PhonemesMappingRequest) = copy(
-    notes = notes.mapNotNull { note -> note.replacePhonemes(request).takeIf { it.lyric.isNotEmpty() } }
+fun Track.replacePhonemes(request: PhonemesMappingRequest?) = copy(
+    notes = notes.mapNotNull { note -> note.replacePhonemes(request) }
         .validateNotes(),
 )
 
-fun Note.replacePhonemes(request: PhonemesMappingRequest): Note {
+fun Note.replacePhonemes(request: PhonemesMappingRequest?): Note {
     val input = phoneme?.split(" ") ?: return this
+    if (request == null) return copy(phoneme = null)
     val output = mutableListOf<String>()
     var pos = 0
     while (pos <= input.lastIndex) {

src/jsMain/kotlin/ui/ConfigurationEditor.kt

@@ -30,6 +30,8 @@ import process.lyrics.chinese.convertChineseLyricsToPinyin
 import process.lyrics.japanese.convertJapaneseLyrics
 import process.lyrics.mapLyrics
 import process.lyrics.replaceLyrics
+import process.phonemes.PhonemesMappingRequest
+import process.phonemes.mapPhonemes
 import process.projectZoomFactorOptions
 import process.zoom
 import react.ChildrenBuilder
@@ -50,6 +52,7 @@ import ui.configuration.ChinesePinyinConversionBlock
 import ui.configuration.JapaneseLyricsConversionBlock
 import ui.configuration.LyricsMappingBlock
 import ui.configuration.LyricsReplacementBlock
+import ui.configuration.PhonemesConversionBlock
 import ui.configuration.PitchConversionBlock
 import ui.configuration.ProjectSplitBlock
 import ui.configuration.ProjectZoomBlock
@@ -170,6 +173,31 @@ val ConfigurationEditor = scopedFC<ConfigurationEditorProps> { props, scope ->
                 ?: FeatureConfig.SplitProject.getDefault(props.outputFormat).maxTrackCount.toString(),
         )
     }
+    val (phonemesConversion, setPhonemesConversion) = useState {
+        currentConfigs?.phonemesConversion?.let {
+            return@useState it
+        }
+
+        val hasPhonemes = props.projects.any { Feature.ConvertPhonemes.isAvailable(it) }
+        val isAvailable = hasPhonemes &&
+            props.outputFormat.availableFeaturesForGeneration.contains(Feature.ConvertPhonemes)
+        // Enable for VOCALOID -> VOCALOID or UfData by default
+        val enabledFormats = Format.vocaloidFormats.plus(Format.UfData)
+        val isOn = isAvailable && (
+            (
+                props.projects.all { it.format in enabledFormats } &&
+                    props.outputFormat in enabledFormats
+                ) ||
+                props.projects.all { it.format == props.outputFormat }
+            )
+        PhonemesConversionState(
+            isAvailable = isAvailable,
+            isOn = isOn,
+            useMapping = false,
+            mappingPresetName = null,
+            mappingRequest = PhonemesMappingRequest(),
+        )
+    }
     var dialogError by useState(DialogErrorState())
 
     fun isReady() = listOf(
@@ -206,6 +234,10 @@ val ConfigurationEditor = scopedFC<ConfigurationEditorProps> { props, scope ->
         initialState = lyricsMapping
         submitState = setLyricsMapping
     }
+    if (phonemesConversion.isAvailable) PhonemesConversionBlock {
+        initialState = phonemesConversion
+        submitState = setPhonemesConversion
+    }
     SlightRestsFillingBlock {
         initialState = slightRestsFilling
         submitState = setSlightRestsFilling
@@ -235,6 +267,7 @@ val ConfigurationEditor = scopedFC<ConfigurationEditorProps> { props, scope ->
         pitchConversion,
         projectZoom,
         projectSplit,
+        phonemesConversion,
         setProgress = { progress = it },
         onDialogError = { dialogError = it },
     )
@@ -261,6 +294,7 @@ private fun ChildrenBuilder.buildNextButton(
     pitchConversion: PitchConversionState,
     projectZoom: ProjectZoomState,
     projectSplit: ProjectSplitState,
+    phonemesConversion: PhonemesConversionState,
     setProgress: (ProgressProps) -> Unit,
     onDialogError: (DialogErrorState) -> Unit,
 ) {
@@ -284,6 +318,7 @@ private fun ChildrenBuilder.buildNextButton(
                     pitchConversion,
                     projectZoom,
                     projectSplit,
+                    phonemesConversion,
                     setProgress,
                     onDialogError,
                 )
@@ -304,6 +339,7 @@ private fun process(
     pitchConversion: PitchConversionState,
     projectZoom: ProjectZoomState,
     projectSplit: ProjectSplitState,
+    phonemesConversion: PhonemesConversionState,
     setProgress: (ProgressProps) -> Unit,
     onDialogError: (DialogErrorState) -> Unit,
 ) {
@@ -335,6 +371,7 @@ private fun process(
                     .runIf(projectZoom.isOn) {
                         zoom(projectZoom.factorValue)
                     }
+                    .mapPhonemes(phonemesConversion.resolvedMappingRequest)
 
                 val featureConfigs = buildList {
                     if (pitchConversion.isOn) add(FeatureConfig.ConvertPitch)
@@ -358,6 +395,7 @@ private fun process(
                 pitchConversion,
                 projectZoom,
                 projectSplit,
+                phonemesConversion,
             )
             window.localStorage.setItem("currentConfigs", json.encodeToString(currentConfigs))
 
@@ -453,6 +491,34 @@ data class LyricsMappingState(
     }
 }
 
+@Serializable
+data class PhonemesConversionState(
+    val isAvailable: Boolean,
+    val isOn: Boolean,
+    val useMapping: Boolean,
+    val mappingPresetName: String?,
+    val mappingRequest: PhonemesMappingRequest,
+) : SubState() {
+    override val isReady: Boolean get() = if (isAvailable && isOn && useMapping) mappingRequest.isValid else true
+
+    val resolvedMappingRequest: PhonemesMappingRequest?
+        get() = if (isOn) {
+            if (useMapping) {
+                mappingRequest
+            } else {
+                PhonemesMappingRequest("")
+            }
+        } else {
+            null
+        }
+
+    fun updatePresetName() = when {
+        mappingPresetName == null -> this
+        PhonemesMappingRequest.findPreset(mappingPresetName) == mappingRequest -> this
+        else -> copy(mappingPresetName = null)
+    }
+}
+
 @Serializable
 data class SlightRestsFillingState(
     val isOn: Boolean,
@@ -502,4 +568,5 @@ data class Configs(
     val pitchConversion: PitchConversionState,
     val projectZoom: ProjectZoomState,
     val projectSplit: ProjectSplitState,
+    val phonemesConversion: PhonemesConversionState,
 )