VoiceStudio is the open-source, fully-local ElevenLabs alternative — voice cloning, voice design, video dubbing, dictation, transcription & audiobook creation in 646 languages.
Python aiaudiobookcudadubbingelevenlabs-alternativehuggingfacelocal-firstmlxomnivoice-studiospeech-to-texttext-to-speechtranscriptiontranslatettsvoice-aivoice-cloningvoice-generationvoicestudioworkflow
fix(mobile): spell sentence-final CJK punctuation as escapes and correct the timeline gap test expectation
debpalashf345e020a6ed
diff --git a/mobile/android/core/local/src/main/kotlin/com/voicestudio/mobile/local/LocalEngine.kt b/mobile/android/core/local/src/main/kotlin/com/voicestudio/mobile/local/LocalEngine.ktindex e9bd33f21..154169ed9 100644--- a/mobile/android/core/local/src/main/kotlin/com/voicestudio/mobile/local/LocalEngine.kt+++ b/mobile/android/core/local/src/main/kotlin/com/voicestudio/mobile/local/LocalEngine.kt@@ -219,7 +219,7 @@ internal fun splitUtterances(text: String,tokens: List<String>,timestamps: List<val groups=mutableListOf(mutableListOf(0))for(i in 1 until tokens.size) {val gap=timestamps[i]-(if(durations.size==tokens.size) endOf(i-1) else timestamps[i-1].toDouble())- val sentenceEnd=tokens[i-1].trimEnd().lastOrNull()?.let {it in ".!?。!?"}==true+ val sentenceEnd=tokens[i-1].trimEnd().lastOrNull()?.let {it in ".!?\u3002\uFF01\uFF1F"}==trueif(gap>=UTTERANCE_GAP||(sentenceEnd&&gap>=SENTENCE_GAP)) groups+=mutableListOf(i) else groups.last()+=i}val pieces=groups.mapIndexedNotNull {n,g ->diff --git a/mobile/android/core/local/src/test/kotlin/com/voicestudio/mobile/local/DubConcurrencyAndTimingTest.kt b/mobile/android/core/local/src/test/kotlin/com/voicestudio/mobile/local/DubConcurrencyAndTimingTest.ktindex 35f694e08..8628aaef3 100644--- a/mobile/android/core/local/src/test/kotlin/com/voicestudio/mobile/local/DubConcurrencyAndTimingTest.kt+++ b/mobile/android/core/local/src/test/kotlin/com/voicestudio/mobile/local/DubConcurrencyAndTimingTest.kt@@ -41,7 +41,7 @@ class DubConcurrencyAndTimingTest {@Test fun rawTimelineInsertsSilenceForLateStartAndGaps() {assertEquals(24000L,timelineShift(1_000_000,0))- assertEquals(12000L,timelineShift(1_500_000,12000))+ assertEquals(24000L,timelineShift(1_500_000,12000)) // 1.5 s = 36000 samples, 12000 written}@Test fun rawTimelineDropsOverlapAndIgnoresJitterAndMissingTime() {assertEquals(-2400L,timelineShift(900_000,24000))diff --git a/mobile/ios/VoiceStudioKit/Sources/VoiceStudioKit/Local/LocalEngine.swift b/mobile/ios/VoiceStudioKit/Sources/VoiceStudioKit/Local/LocalEngine.swiftindex e521a9890..f74fccf88 100644--- a/mobile/ios/VoiceStudioKit/Sources/VoiceStudioKit/Local/LocalEngine.swift+++ b/mobile/ios/VoiceStudioKit/Sources/VoiceStudioKit/Local/LocalEngine.swift@@ -76,7 +76,7 @@ func splitUtterances(text: String,tokens: [String],timestamps: [Double],ends: [Dvar groups: [[Int]] = [[0]]for i in 1..<max(1,tokens.count) {let gap = timestamps[i]-endOf(i-1)- let sentenceEnd = tokens[i-1].trimmingCharacters(in: .whitespaces).last.map { ".!?。!?".contains($0) } ?? false+ let sentenceEnd = tokens[i-1].trimmingCharacters(in: .whitespaces).last.map { ".!?\u{3002}\u{FF01}\u{FF1F}".contains($0) } ?? falseif gap >= utteranceGap || (sentenceEnd && gap >= sentenceGap) { groups.append([i]) } else { groups[groups.count-1].append(i) }}var pieces: [SpeechSegment] = []