Meta Launches Muse Voice Transcribe With Real-Time Multilingual Transcription
· Technology · Engadget, Indian Express
Meta has released Muse Voice Transcribe, its first real-time audio transcription model from Meta Superintelligence Lab (MSI). The model can distinguish between more than 20 speakers in a single session, switch seamlessly between languages mid-sentence, and handle code-switching where sentences mix words from multiple languages. It is trained on over 70 languages, with 25 validated at launch, and uses adaptive delay to wait longer on difficult words while committing faster on easy ones. Meta is pricing the model at $3 per 1,000 audio minutes through Muse Code and its Model API, with dictation already live in the Meta desktop app. The release comes less than a week after Google launched Gemini 3.5 Transcribe with comparable capabilities.
Why it matters
Real-time multilingual transcription with speaker diarization is a core capability for meeting tools, accessibility services, and global enterprise workflows, putting Meta in direct competition with Google's audio stack and OpenAI's Whisper-class models. Pricing at $3 per 1,000 audio minutes sets an aggressive benchmark for developers building voice products.
Read the original report — Engadget
Join us on Telegram
Breaking news the moment it lands. At 10,000 members we ship the Android app.