Chariot Launches Multilingual TTS Model Supporting 11 Indian Languages and English

Chariot Unveils Advanced Multilingual TTS Model

Chariot, an Indian AI research lab founded by former Google DeepMind and Meta FAIR researchers, has launched Chariot-1-TTS-Multilingual, a text-to-speech model supporting 11 Indian languages and English.1

1

The Delhi-headquartered company, previously known as Gan.ai, was founded by Parth Sarthi and Suvrat Bhooshan.1

1

This multilingual speech model supports Hindi, Marathi, Bengali, Tamil, Telugu, Kannada, Malayalam, Gujarati, Punjabi, Odia, and Indian English.1

1

Contextual Reasoning Sets Model Apart

Unlike conventional TTS systems relying on text normalisation rules, Chariot's model uses contextual reasoning for pronunciation to determine how written expressions should be spoken.1

1

The model distinguishes context-dependent meanings, such as differentiating 'B+' referring to a blood group versus a grade.2

2

It handles contextual pronunciation of currency amounts, dates, abbreviations, and technical expressions.1

1

"People use numbers, symbols and abbreviations, and move between languages in the same sentence," Parth Sarthi explained. "A speech model has to understand what those choices mean and how to say them."2

2

Comprehensive Voice Capabilities and Performance

The Chariot-1-TTS-Multilingual model supports more than 50 voices and enables code-mixing within sentences between any two supported languages.1

1

It offers instant voice cloning using just a few seconds of reference audio and can generate speech from up to 1,000 characters in a single request.2

2

Chariot reports the model ranks first across several measures in comparisons with selected models from ElevenLabs, Cartesia, Microsoft, OpenAI, AWS, and Sarvam.1

1

The model achieved a character error rate of 7.60% for Hindi and 6.47% across nine other Indic languages.2

2

On an English verbalisation benchmark covering more than 1,000 sentences across 40 categories, it recorded 93.78% accuracy.1

1

Deployment for Voice Agents and Customer Service

The model targets applications including voice agents, customer service systems, educational platforms, and multilingual public helplines.2

2

Banks can deploy it for payment reminders, learning platforms for narration, and public helplines for spoken information across languages.2

2

Developers can access the voice catalogue or create custom voices using reference recordings.1

1

Chariot showcases projects with brands including Zomato, Doordarshan, Coca-Cola, Cadbury, and HDFC Bank, demonstrating applications from personalised content to multilingual broadcasts.1

1

Building Foundation Models from India

Suvrat Bhooshan stated: "For voice technology to be useful in India, it needs to work with the languages people speak and the way they move between them. Our ambition is to build foundation models that understand, generate and reason in speech."2

2

The company recently launched Chariot-1-ASR Realtime, a speech recognition model that reportedly outperformed every streaming and batch transcription system tested across 10 Indian languages.1

1

Chariot is among 12 organisations selected under the IndiaAI Mission's Foundation Models pillar, launched in March 2024 with an allocation of ₹10,372 crore to support indigenous AI infrastructure and models.1

1

This positions the AI lab to compete globally while addressing India's linguistic diversity, potentially reshaping how multilingual voice technology serves diverse populations.