$VibeVoice
Microsoft released an open-source voice AI system called VibeVoice that can transcribe up to 60 minutes of audio in one pass with speaker identification, timing, and full context preserved