Meta just beat OpenAI and Google at real-time transcription
Meta’s Superintelligence Labs on Tuesday launched Muse Voice Transcribe, a new real-time speech recognition model that, at least on some The post Meta just beat OpenAI and Google at real-time transcription appeared first on The New Stack .
Meta's Superintelligence Labs has unveiled Muse Voice Transcribe, a cutting-edge real-time speech recognition model that outperforms competitors like OpenAI, Google, and ElevenLabs in key benchmarks. This marks Meta's first "real-time audio perception model." Muse Voice Transcribe can recognize over 20 speakers and supports 70 languages, including multi-lingual conversations.
The model's latency is controlled through an "adaptive delay" mechanism, ensuring accurate transcription of difficult words while quick transcription of easier ones. Priced at $3.00 per 1,000 audio-minutes, Muse Voice Transcribe is available through the Meta Model API, Meta AI for Mac, and Muse Code. While not open source, the model demonstrates superior performance in English speech recognition word error rate (3.1%) compared to competitors.
Written by urgent.news from The New Stack's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
