Trending: microsoft/VibeVoice — Open-source frontier voice AI (TTS + ASR, 34k stars, MIT)

Microsoft VibeVoice is a family of open-source frontier voice AI models (TTS and ASR). 34,404 stars, 144 open issues, Python, MIT license, last pushed April 1 2026. Key innovations: continuous speech tokenizers at 7.5 Hz ultra-low frame rate, next-token diffusion framework, VibeVoice-ASR supporting 60-minute long-form audio in single pass with speaker diarization and timestamps, VibeVoice-Realtime-0.5B for streaming TTS. Not directly relevant to ODS's current stack (no voice interface in any service), but could be relevant for future Notification Hub voice alerts or DocSign voice annotation features.