Overview
Chariot is a Delhi-based frontier AI research lab and one of four teams selected under India's IndiaAI Mission to build national foundation models. Its thesis is that speech should be treated as a first-class modality (like text) rather than a pipeline of separate transcription/synthesis/dubbing tools, with the long-term goal of building "speech reasoning" models that understand meaning, delivery, timing, and action in one system. Its first public products—Chariot-1 TTS and Chariot-1-ASR Realtime—are live in English and multiple Indian languages.
Key Features
Context-aware text-to-speech (Chariot-1 TTS): Understands what text is trying to say, not just how to pronounce tokens—handling numbers, abbreviations, dates, currencies, URLs, and units naturally without forcing users to expand or spell things out.
Real-time speech recognition (Chariot-1-ASR Realtime): Supports 10 Indian languages and claims to beat every system tested, including ones that wait for full recordings.
Multilingual TTS with 50+ voices and voice cloning: Speaks 11 Indic languages plus English (American, British, Indian accents), with code-mixing, instant voice cloning, and context-aware pronunciation.
Speech reasoning roadmap: Building toward models that reason across meaning, delivery, timing, and action—handling interruption, pause, uncertainty, emotion, and task completion natively in speech rather than via a transcript-to-LLM-to-TTS pipeline.
Use Cases
Developers can generate and stream speech via API for products needing natural, context-aware voice output.
Enterprises can use it for customer service, technical/scientific narration, and multilingual content (claimed #1 in accuracy on hard examples for customer service and technical/scientific text).
Media/entertainment: The team has shipped high-profile projects—Mann Ki Baat dubbed in the PM's voice across languages, a Shah Rukh Khan cricket Q&A bot, 600,000 personalized birthday songs, and Zomato celebrity-voiced ads.