Meta Superintelligence Labs has announced Muse Voice Transcribe, its first real-time speech recognition model. The company describes this as a major milestone for Meta's real-time voice models.
Model ReleasesMetaMeta Superintelligence LabsMuse Voice Transcribe
Meta Announces First Real-Time Speech Recognition Model Muse Voice Transcribe
This article is a translation. Read the Japanese original
The model features real-time streaming speech recognition capabilities. It also includes speaker diarization supporting more than 20 speakers and end-of-speech detection.
According to the AI evaluation platform "Artificial Analysis," the model has secured the number one spot in its rankings.
Source: Metaがリアルタイム文字起こしAI「Muse Voice Transcribe」をリリース (GIGAZINE, 2026-09-02)