NEW: Meta introduces Muse Voice Transcribe, a real-time audio perception model from Meta Superintelligence Labs that delivers streaming speech-to-text and speaker diarization with 20+ speakers.