KittenTTS ships several model sizes. Pick the smallest model that sounds good enough for your use case.
Start with nano-int8 when app size and download size matter. Use mini when
quality is more important.
| Model | ID | Parameters | Approx download | Hugging Face |
|---|---|---|---|---|
| Nano int8 | 'nano-int8' |
15M | 25 MB | kitten-tts-nano-0.8-int8 |
| Nano fp32 | 'nano' |
15M | 56 MB | kitten-tts-nano-0.8 |
| Micro | 'micro' |
40M | 41 MB | kitten-tts-micro-0.8 |
| Mini | 'mini' |
80M | 80 MB | kitten-tts-mini-0.8 |
const tts = await KittenTTS.create({
model: 'nano-int8',
});| Voice | ID | Character |
|---|---|---|
| Bella | 'bella' |
Warm and expressive |
| Jasper | 'jasper' |
Clear and conversational |
| Luna | 'luna' |
Calm and smooth |
| Bruno | 'bruno' |
Deep and steady |
| Rosie | 'rosie' |
Bright and friendly |
| Hugo | 'hugo' |
Authoritative |
| Kiki | 'kiki' |
Lively and energetic |
| Leo | 'leo' |
Relaxed and natural |
await tts.speak('Luna speaking.', { voice: 'luna' });
await tts.speak('Slower Bella speaking.', { voice: 'bella', speed: 0.8 });