Meta AI: Real-time Transcription for 5 Indian Languages

Meta has launched Muse Voice Transcribe, an AI model that transcribes live speech. This new tool supports five Indian languages natively. Here's the thing: Muse Voice Transcribe can identify and separate over 20 speakers in recordings longer than an hour. It also handles conversations where people switch between languages. The model was trained on audio from more than 70 languages, with 25 validated at launch. For India, it supports Hindi, Tamil, Telugu, Kannada, and Malayalam. What's interesting is that it ranked first on a streaming speech-to-text leaderboard as of September 1, 2026. It achieved a 3.1% word error rate. This model can adapt its processing delay, quickly outputting easy words and taking more time for difficult ones. The bottom line: This technology aims to make real-time transcription faster and more accurate for complex, multi-speaker, and multi-lingual conversations.