accompanist-lyrics-phonetics

Accompanist includes a group of libraries:

• lyrics-core — Parse lyrics files, retain timing and metadata, and export to other formats.
• lyrics-ui — Display synchronized lyrics with Compose Multiplatform.
• lyrics-phonetics — Resolve pronunciation on device and format captions for lyrics.

This repository hosts the lyrics-phonetics code, a Kotlin Multiplatform library for Android, JVM and iOS.

• 🌏 Multilingual Readings: Mainland and Taiwan Mandarin, Hong Kong Cantonese, and Japanese.
• 🧠 Contextual Pronunciation: Language hints, alternative readings and custom overrides.
• 🎤 Lyrics Alignment: Project pronunciation onto original lyric ranges.
• 📱 Offline & Multiplatform: Optional data packs for Android, JVM and iOS.

Add the runtime and the data packs you need to your build.gradle.kts:

dependencies {
    implementation("com.mocharealm.accompanist:lyrics-phonetics-runtime:0.1.0")
    implementation("com.mocharealm.accompanist:lyrics-phonetics-data-mandarin:0.1.0")
    implementation("com.mocharealm.accompanist:lyrics-phonetics-data-cantonese:0.1.0")
    implementation("com.mocharealm.accompanist:lyrics-phonetics-data-japanese:0.1.0")
}

For Kotlin Multiplatform, add these to commonMain.dependencies. Include only the language packs you need.

On Android, keep dictionary assets uncompressed so the loader can map them:

android {
    androidResources { noCompress += "lpd" }
}

On iOS, copy the data packs and their licenses from the resources ZIP into the app bundle.

import com.mocharealm.accompanist.lyrics.phonetics.*
import com.mocharealm.accompanist.lyrics.phonetics.data.*

// JVM example. On Android use AndroidPackLoader(context);
// on iOS use IosBundlePackLoader().
val loader = SharedPackLoader(ClasspathPackLoader())
val engine = PhoneticEngine(listOf(
    MandarinData.mainland(loader),
    MandarinData.taiwan(loader),
    CantoneseData.hongKong(loader),
    JapaneseData.japanese(loader),
))
val result = engine.parse(
    "银行行走",
    ParseOptions(defaultLocale = ReadingLocale.MANDARIN_CN),
)

println(AsciiFormatter().format(result).latin) // yin hang xing zou
println(AsciiFormatter(FormatOptions(tones = ToneStyle.NUMBERS)).format(result).latin)
// yin2 hang2 xing2 zou3

• Dictionary format
• Data sources and licenses
• Benchmarks

Contributions are welcome! Please open an issue before proposing major changes.

Code is licensed under Apache License 2.0. See LICENSE. Language data has separate licenses; see data sources and licenses and retain the bundled notices.