API reference

Construction

KittenTTS.create(options?, onProgress?): Promise<KittenTTS>

Instance APIs

Method	Purpose
`generate(text, options?)`	Return one `KittenTTSResult`
`stream(text, options?)`	Async generator of sentence results
`speak(text, options?)`	Generate and play
`play(result, options?)`	Play an existing result
`createPlaybackQueue()`	FIFO playback queue
`stopSpeaking()` / `stop()`	Stop active audio
`pauseSpeaking()`	Pause supported player
`resumeSpeaking()`	Resume supported player
`isSpeaking`	Current playback state
`dispose()`	Release engine and phonemizer resources

`KittenTTSResult`

samples, sampleRate, duration, voice, effectiveSpeed, inputText, wordTimings, wavData(), wavBase64(), mp3Data(options?), and mp3Base64(options?).

Cache APIs

cacheInfo, predownload, validateAssets, redownloadModel, and clearModelCache are static.

Public support types

The package exports model and voice helpers, progress and cache types, AssetStorage implementations, CEPhonemizer, MP3Encoder, PlaybackQueue, WAVEncoder, AudioPlayer, and AudioPlayOptions.

Errors

Use isKittenTTSError(error) and inspect KittenTTSErrorCode. The error family covers blank input, downloads, invalid assets, phonemization, inference, and playback.

The generated lib/index.d.ts from the pinned source commit is authoritative for exact signatures.

Construction​

Instance APIs​

KittenTTSResult​

Cache APIs​

Public support types​

Errors​