Fix UTAU pitch conversion for multiple key points that are close to e…

…ach other (#180) * Fix UTAU/OpenUtau pitch import for very close pitch points * Fix resampling during ustx pitch importing
sdercolin · Jul 29, 2024 · 17de7cb · 17de7cb
1 parent 9cad76e
commit 17de7cb
Show file tree

Hide file tree

Showing 3 changed files with 32 additions and 4 deletions.
diff --git a/core/src/main/kotlin/core/process/pitch/OpenUtauPitchConversion.kt b/core/src/main/kotlin/core/process/pitch/OpenUtauPitchConversion.kt
@@ -11,6 +11,10 @@ import kotlin.math.roundToInt
 
 private const val SAMPLING_INTERVAL_TICK = 5L
 
+// Not smaller than the sampling interval of the pitch data used in pitch generation targeting any format.
+// This is used for placing two key points far enough to ensure they are not merged by the resampling process.
+private const val SAFE_SAMPLING_INTERVAL_TICK = 5L
+
 data class OpenUtauNotePitchData(
     val points: List<Point>,
     val vibrato: UtauNoteVibratoParams,
@@ -44,12 +48,17 @@ fun pitchFromUstxPart(notes: List<Note>, pitchData: OpenUtauPartPitchData, tempo
     // Extract pitch points from notes
     val notePointsList = mutableListOf<List<Pair<Long, Double>>>()
     val tickTimeTransformer = TickTimeTransformer(tempos)
+    var lastKeyPos = -SAFE_SAMPLING_INTERVAL_TICK
     for ((note, notePitch) in notes.zip(pitchData.notes)) {
         val points = mutableListOf<Pair<Long, Double>>()
         var lastPointShape = OpenUtauNotePitchData.Shape.EaseInOut
         val noteStartInMillis = tickTimeTransformer.tickToMilliSec(note.tickOn)
+        val keyPointPositions = mutableListOf<Long>()
         for (rawPoint in notePitch.points) {
             val x = tickTimeTransformer.milliSecToTick(noteStartInMillis + rawPoint.x)
+                .coerceAtLeast(lastKeyPos + SAFE_SAMPLING_INTERVAL_TICK)
+            lastKeyPos = x
+            keyPointPositions.add(x)
             val y = rawPoint.y / 10
             val thisPoint = x to y
             val lastPoint = points.lastOrNull()
@@ -67,7 +76,8 @@ fun pitchFromUstxPart(notes: List<Note>, pitchData: OpenUtauPartPitchData, tempo
         val pointsInNoteWithVibrato = pointsIn
             .appendUtauNoteVibrato(notePitch.vibrato, note, tickTimeTransformer, SAMPLING_INTERVAL_TICK)
         val pointsWithVibrato = pointsBefore + pointsInNoteWithVibrato + pointsAfter
-        notePointsList.add(pointsWithVibrato.resampled(SAMPLING_INTERVAL_TICK))
+        val pointsResampled = pointsWithVibrato.resampled(SAMPLING_INTERVAL_TICK, keyPointPositions)
+        notePointsList.add(pointsResampled)
     }
 
     // Divide note with points into sections,
@@ -243,10 +253,18 @@ private fun MutableList<Pair<Long, Double>>.appendStartAndEndPoint(note: Note) {
     }
 }
 
-private fun List<Pair<Long, Double>>.resampled(interval: Long): List<Pair<Long, Double>> =
+private fun List<Pair<Long, Double>>.resampled(
+    interval: Long,
+    keyPointPositions: List<Long> = emptyList(),
+): List<Pair<Long, Double>> =
     groupBy { it.first / interval * interval }
         .map { (mergedTick, points) ->
-            mergedTick to points.map { it.second }.average()
+            val keyPoint = points.find { it.first in keyPointPositions }
+            if (keyPoint != null) {
+                mergedTick to keyPoint.second
+            } else {
+                mergedTick to points.map { it.second }.average()
+            }
         }
         .sortedBy { it.first }
         .fold(listOf()) { acc, point ->

diff --git a/core/src/main/kotlin/core/process/pitch/UtauMode2PitchConversion.kt b/core/src/main/kotlin/core/process/pitch/UtauMode2PitchConversion.kt
@@ -13,6 +13,11 @@ import core.util.runIf
 
 private const val SAMPLING_INTERVAL_TICK = 4L
 
+// Not smaller than the sampling interval of the pitch data used in pitch generation targeting any format.
+// This is used for placing two key points far enough to ensure they are not merged by the resampling process.
+// See also: core.process.pitch.OpenUtauPitchConversion.kt
+private const val SAFE_SAMPLING_INTERVAL_TICK = 5L
+
 data class UtauMode2TrackPitchData(
     val notes: List<UtauMode2NotePitchData?>,
 )
@@ -87,12 +92,15 @@ fun pitchFromUtauMode2Track(pitchData: UtauMode2TrackPitchData?, notes: List<Not
     val pitchPoints = mutableListOf<Pair<Long, Double>>()
     var lastNote: Note? = null
     var pendingPitchPoints = listOf<Pair<Long, Double>>()
+    var lastKeyPos = -SAFE_SAMPLING_INTERVAL_TICK
     for ((note, notePitch) in notePitches) {
         val points = mutableListOf<Pair<Long, Double>>()
         val noteStartInMillis = tickTimeTransformer.tickToMilliSec(note.tickOn)
         if (notePitch?.start != null) {
             var posInMillis = noteStartInMillis + notePitch.start
             var tickPos = tickTimeTransformer.milliSecToTick(posInMillis)
+                .coerceAtLeast(lastKeyPos + SAFE_SAMPLING_INTERVAL_TICK)
+            lastKeyPos = tickPos
             val startShift =
                 if (note.tickOn == lastNote?.tickOff) {
                     // always same value as the last note
@@ -108,6 +116,8 @@ fun pitchFromUtauMode2Track(pitchData: UtauMode2TrackPitchData?, notes: List<Not
                 val curveType = notePitch.curveTypes.getOrNull(index) ?: ""
                 posInMillis += width
                 tickPos = tickTimeTransformer.milliSecToTick(posInMillis)
+                    .coerceAtLeast(lastKeyPos + SAFE_SAMPLING_INTERVAL_TICK)
+                lastKeyPos = tickPos
                 val thisPoint = tickPos to (shift / 10)
                 val lastPoint = points.last()
                 if (thisPoint.second != lastPoint.second) {

diff --git a/src/jsMain/kotlin/Main.kt b/src/jsMain/kotlin/Main.kt
@@ -6,7 +6,7 @@ import ui.strings.Language
 import ui.strings.initializeI18n
 
 const val APP_NAME = "UtaFormatix"
-const val APP_VERSION = "3.22"
+const val APP_VERSION = "3.22.1"
 
 suspend fun main() {
     initializeI18n(Language.English)