Sarvam AI launches Saaras V4 transcribing Indian languages and English
Technology
Sarvam AI just launched Saaras V4, a new speech recognition model that can listen to and transcribe conversations in multiple Indian languages as well as English, even in noisy, real-world settings.
Built with an audio encoder with a 3-billion-parameter hybrid state-space language model developed by Sarvam AI, it's designed to handle everything from mixed languages to different accents.
Saaras V4 supports 5 transcript styles
Saaras V4 gives you transcripts in five styles, such as verbatim or code-mixed, and can handle long audio or live streaming without breaking a sweat.
If you're into building apps, the model is ready for action via APIs (Python and Node.js) and works with platforms like Vercel AI SDK and LiveKit Agents.