rendchevi/nix-tts

🐤 Nix-TTS: Lightweight and End-to-end Text-to-Speech via Module-wise Distillation

50
/ 100
Established

Module-wise knowledge distillation compresses a large non-autoregressive end-to-end teacher model into a 5.23M-parameter student achieving 11.9× real-time speedup on CPU and 8.36× on Raspberry Pi while maintaining voice quality. The vocoder-free architecture enables independent distillation of encoder and decoder components, allowing flexible optimization for resource-constrained deployment. Pretrained models are available in ONNX format alongside a Hugging Face demo and Python inference API.

261 stars.

No Package No Dependents
Maintenance 6 / 25
Adoption 10 / 25
Maturity 16 / 25
Community 18 / 25

How are scores calculated?

Stars

261

Forks

35

Language

Python

License

MIT

Last pushed

Nov 15, 2025

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/voice-ai/rendchevi/nix-tts"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.