New Delhi: An IndiaAI Mission-backed startup has launched a real-time speech recognition model it says beats the best global and domestic systems in 10 Indian languages, even as the speech AI market slides into a price war.
Chariot's new offering, Chariot-1-ASR Realtime, is claimed to outperform specialised speech recognition models, including those from Google, ElevenLabs and Indian AI developers, across publicly available benchmarks. The company says it handles regional accents, background noise and code-mixed speech - the everyday reality of how Indians talk.
"Currently, it is a race to the bottom. We see every voice AI company also trying to build their own ASR systems because they think the margins they can get perhaps are more if they build it in-house," founder and CEO Parth Sarthi said.
Sarthi said the company evaluated its model against leading commercial streaming and batch transcription systems across 79 public datasets, including benchmarks designed to test speech recognition in challenging conditions. "We decided that we're not going to compare our streaming model just with other streaming models," he said. It wanted to measure up against the strongest available system for each language.
Using NVIDIA's NeMo framework, Chariot is developing an 8-billion-parameter model for real-time text-to-speech, aimed at accessibility and digital interactions across consumer and enterprise use cases.
Sarthi studied computer science at Stanford University and worked at Google DeepMind. He said the company's priority remains building and improving its underlying models rather than aggressively acquiring enterprise customers.
Speech recognition sits at the heart of government schemes, health services, customer support and education. A strong homegrown option for Indian languages could cut reliance on foreign models, experts say.
Competition in speech technology has intensified globally, and pricing has become a decisive factor as companies hunt for cheaper, accurate models. Handling code-mixed speech is especially hard in a country as multilingual as India. The government's IndiaAI Mission has placed a special focus on language technology. Such tools can deliver farm advisories, health information and banking services in local languages. Experts say accurate speech recognition is foundational to the next phase of AI adoption.