Voices - RunAnywhere Documentation

val voices = RunAnywhere.tts.voices()

Signature

suspend fun voices(): List<Voice>

The list reflects the loaded speech model, so call it after the model is in place rather than at launch. Load a different TTS model and the list changes.

Selecting one

val handle = RunAnywhere.tts.speak(
    "Hello",
    TtsOptions(voice = "en_US-amy-medium")
)

With voice unset, the model’s own default is used.

Language and voice are separate

language sets the language code; voice picks a speaker. A multi-voice model exposes several speakers for one language, and a single-voice model ignores voice entirely.

TtsOptions(
    voice = "en_US-amy-medium",
    language = "en-US",
    speed = 1.0f,
    pitch = 1.0f
)

Letting a person choose

class VoicePicker : ViewModel() {
    private val _voices = MutableStateFlow<List<Voice>>(emptyList())
    val voices = _voices.asStateFlow()

    var selected: String? = null

    fun load() {
        viewModelScope.launch {
            _voices.value = RunAnywhere.tts.voices()
        }
    }

    fun preview() {
        val voice = selected ?: return
        viewModelScope.launch {
            RunAnywhere.tts.speak("This is how I sound.", TtsOptions(voice = voice))
        }
    }
}