Fast, small, open Kannada TTS:
Kokoro-82M
(distilled StyleTTS2) fine-tuned for Kannada
with a trained diffusion prosody sampler —
the component stock Kokoro drops, and the difference between robotic and natural prosody.
Use the inference script (handles the fork's architecture code + diffusion sampling + best-of-N):
1git clone https://github.com/sammy4321/Kokoro-Indic-Fine-Tuning
2cd Kokoro-Indic-Fine-Tuning && pip install -r demo/requirements.txt huggingface_hub
3python inference/synthesize.py --text "ನಮಸ್ಕಾರ!" --out out.wav --best-of 3
82M params · 24 kHz · trained on ~11 h public Kannada speech
(
SPRINGLab IndicTTS +
AI4Bharat Rasa, 17 speakers) for <$100 of
compute. ASR round-trip CER 0.00 on clean text. Known gaps: numerals-as-digits and heavy
English code-switching need a text-normalization frontend (roadmap).
License: code Apache-2.0; weights are derived from the datasets above — check their
terms before commercial use; attribution required. Built on StyleTTS2 (MIT), Kokoro-82M
(Apache-2.0), and the
kikiri-tts recipe.
