← Back to feed

VibeVoice

GitHub Repo Pretty sure · TTS removal is honest.
https://github.com/microsoft/VibeVoice

Microsoft's speech AI family that actually solves the 60-minute transcription problem instead of pretending 30-second chunks are sufficient. TTS was pulled for abuse; ASR remains legit.

15%
20%
65%
Slop 15%Signal 20%Science 65%

VibeVoice-ASR is genuine research: 7.5 Hz tokenization is a real efficiency innovation, 60-minute single-pass processing actually addresses a production pain point (context loss in chunked ASR), and the diarization+timestamps output has utility. Paper exists. Models on HF are real code, not smoke. The honesty about pulling TTS for misuse is refreshingly rare in the AI space—no corp slides that under a 'safety review.' Science score reflects actual technical contribution; signal is modest beca...

24646 stars Python 2026-03-27 214 days old

Become a MFer to rate — log in