SILMA-AI/silma-tts

SILMA TTS v1 Official Repo — a Lightweight Open Bilingual Text to Speech Model

42
/ 100
Emerging

Built on the F5-TTS diffusion architecture, this 150M-parameter model achieves ~0.12 real-time factor (RTX 4090) with instant voice cloning and zero-shot style transfer. It includes native Arabic diacritization support via CATT and NeMo text processing for handling both Arabic (Fusha/MSA) and English with full tashkeel awareness. The model is fully compatible with F5-TTS v1.1.7 training pipelines, enabling community fine-tuning while maintaining inference optimizations over the base architecture.

6 stars and 1,118 monthly downloads. Available on PyPI.

Maintenance 13 / 25
Adoption 11 / 25
Maturity 18 / 25
Community 0 / 25

How are scores calculated?

Stars

6

Forks

Language

Python

License

MIT

Last pushed

Mar 15, 2026

Monthly downloads

1,118

Commits (30d)

0

Dependencies

25

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/voice-ai/SILMA-AI/silma-tts"

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