tiny-voca: 4.5M-Parameter TTS Trained Fully on Synthetic Speech
Developer @n4ze3m released tiny-voca, a fully open-source 4.5M-parameter, 18MB English TTS trained entirely on synthetic speech generated by Qwen3-TTS, running on CPU at 70x real-time speed with code, weights, recipes and dataset all public.
2026-11-15 ~ 2026-11-16 · 4 related posts
- tiny-voca: an open 4.5M-param TTS that runs 70x real time on a CPU — nato_nob · 2026-11-15
- tiny-voca: an 18MB, 4.5M-param TTS model trained without any real human voice, runs on CPU — Dan_Jeffries1 · 2026-11-15