Sarvam AI launches Saaras V4 speech recognition model for multilingual voice applications

The new ASR model supports five output modes, 22 Indian languages and low-latency streaming