Microsoft Open-Sources VibeVoice: Frontier Voice AI Family With TTS, ASR, and Realtime Models — 43.6K Stars

GitHub·high signal

Microsoft's VibeVoice is a family of three open-source frontier voice AI models: VibeVoice-ASR-7B (processes 60-minute audio in a single pass, 50+ languages), VibeVoice-TTS-1.5B (generates up to 90 minutes of multi-speaker conversational speech), and VibeVoice-Realtime-0.5B (~300ms first-audible latency streaming TTS). The framework uses continuous speech tokenizers at an ultra-low 7.5 Hz frame rate combined with next-token diffusion. Trending today with 757 stars/day, the TTS model was accepted as an oral presentation at ICLR 2026.