
Orissa Diary
Business News Matrix
Chariot, an AI research lab founded by former Google DeepMind and Meta FAIR researchers, has launched *Chariot-1-ASR Realtime*, a live speech recognition system built on its broader *Chariot-1* foundation model for speech. Unlike most tools built for a single task, Chariot-1 learns speech the way large language models learn text, and is then specialised for jobs like transcription.
The company claims the new model outperforms both real-time and batch transcription systems from Sarvam, Google, ElevenLabs and Deepgram, posting the lowest error rate across all 10 languages tested, including a 23 percent improvement over rivals in Hindi. It also performs strongest on real-world Hindi and Indian-English phone calls, regional accents, and noisy recordings.
Co-founder Parth Sarthi said the model ends the traditional trade-off between "fast, or accurate" transcription. Co-founder Suvrat Bhooshan added that the system reflects Chariot's belief that "speech is one problem, not many," with future releases expected to extend the same foundation to speech generation and beyond.

Sikkim Festival Explores Translation And Oral Traditions

Kannada Classes for Newcomers in Bengaluru This Karunadotsava

CoRover's BharatGPT Indic Language Models Join AIKosh

Delhi Speaker Gupta Bats for Mother Tongue in Education

Gnani AI Launches Prisma v3.0 for Mixed-Language Speech

Chandigarh Akademi Honors Dr. Naresh for Urdu Literature
