Mark Zuckerberg announced Muse Voice Transcribe as Meta Superintelligence Labs' first real-time audio perception model. The system performs streaming speech-to-text, diarization, and endpointing in one model. It uses adaptive delay to balance speed and accuracy. The model supports multiple languages with code-switching. It now powers dictation in the Meta desktop app and voice input in Muse Code. Access is live through the Meta Model API, Meta AI for Mac, and Muse Code, including a zero-data-retention tier on the API. The official research blog post confirms the rollout.