AnalysisAI ModelsNovember 15, 2026

tiny-voca: open 4.5M-parameter TTS runs 70x real time on CPU

Read original source →huggingface.co

The fully open English text-to-speech model has 4,537,291 parameters in an 18 MB package, with two voices, and runs in-browser with nothing sent to a server. Code, weights, and dataset are all released.

2 sources

More stories today

Open the live feed