Sarvam launches Saaras V4 for voice apps across 22 Indian languages

Sarvam AI's Saaras V4 supports multilingual and code-mixed speech through five output modes. Its accuracy and latency figures are company claims that still need independent testing.

Indian artificial intelligence company Sarvam AI has launched Saaras V4, a speech recognition model designed for multilingual voice applications. The company says the system supports 22 Indian languages and English, with automatic language detection and tools for developers building voice products.

Saaras V4 offers five output modes called transcribe, translate, translit, verbatim and codemix. These modes are intended to handle standard transcription, English translation, transliteration, exact........

© News9Live