Models and configuration

Models

ID	Parameters	Approx download
nano-int8	15M	25 MB
nano	15M	56 MB
micro	40M	41 MB
mini	80M	80 MB

Voices

bella, jasper, luna, bruno, rosie, hugo, kiki, and leo.

Configuration

const tts = await KittenTTS.create({
  model: 'nano-int8',
  defaultVoice: 'luna',
  speed: 1.05,
  downloadRetries: 4,
  ortNumThreads: 4,
  maxTokensPerChunk: 400,
  trimTrailingSilence: true,
  silenceThreshold: 0.005,
  maxSilenceTrimMs: 250,
  analytics: false,
});

Option	Default	Purpose
storageDirectory	KittenTTS	Cache namespace or Node root
modelBaseURL	Hugging Face	Model asset mirror
modelFiles	none	Bytes or Node paths
downloadRetries	4	Total attempts per asset
ortNumThreads	4	ONNX intra-op threads
maxTokensPerChunk	400	Long-text inference size
trimTrailingSilence	true	Trim chunk tails
phonemizer	CEPhonemizer	Custom G2P implementation
storage	runtime default	Custom AssetStorage
fetch	globalThis.fetch	Custom transport
ortWasmPath	matching CDN	Self-hosted browser runtime

Speed is clamped to 0.5–2.0; token count is at least 50.