The story’s score and claims

Comments on No. 108 of 122 · huggingface.co · edition of Tuesday 17 November 2026

Tiny Voca: TTS on CPU (45M param, 18MB)

Summary

Muhammed Nazeem has released tiny-voca, an open-source English text-to-speech model with two voices and 4.5 million parameters, trained from scratch on about 18 hours of synthetic speech and running roughly 70 times faster than real time on four CPU threads.

Comments

Ranked by a public rubric, not by votes: what a comment adds counts, not who wrote it. How comments are ranked · Comment rules