Kitten TTS TTS
An ultra-lightweight ONNX model under 80MB that runs high-quality TTS on CPU with no GPU at all.
Kitten TTS by KittenML is built for extreme portability: an ONNX-based model with variants from 15M to 80M parameters, weighing just 25-80 MB on disk. It runs entirely on CPU with 0 VRAM, yet ships 8 built-in voices, adjustable speech speed, and built-in text preprocessing that handles numbers, currencies, and units before synthesis. Output is 24kHz. While its quality sits below the larger models, its tiny size and fast (~2s) CPU inference make it ideal for edge deployment and latency-sensitive applications that can't assume a GPU is present.
At a glance
- Developer
- KittenML
- License
- Apache 2.0
- Tier
- free
- Speed
- fast
- Voice cloning
- No
- Languages
- English
- Max characters
- 5000
Kitten TTS voices
Best for
Fast lightweight TTS, edge deployment, low-latency applications
Kitten TTS TTS — FAQ
Kitten TTS is an ONNX model ranging from 15M to 80M parameters, just 25-80 MB on disk — under 80MB — and it runs on CPU with no GPU required.
It offers 8 built-in voices, adjustable speech speed, 24kHz output, and built-in text preprocessing for numbers, currencies, and units. It is English-only and does not support voice cloning.
Yes. Kitten TTS is Apache 2.0 licensed and sits in the free tier.