- Mayura supports translation between 10 major Indian languages, including Hindi, Tamil, Telugu, and Malayalam.
- Mayura includes a "gender toggle" feature to provide accurate gender representation in first-person translations.
Bengaluru-based startup Sarvam AI has unveiled Mayura, a new translation model aimed at addressing the complexities of English-to-Indic language translation. Unlike traditional models, Mayura focuses on real-world language patterns such as code-mixing and colloquial expressions. It supports 10 major Indian languages, including Hindi, Tamil, Telugu, Malayalam, Punjabi, Odia, Gujarati, Marathi, Kannada, and Bengali.
Sarvam AI developed Mayura using diverse real-world data to ensure that regional dialects, slang, and code-mixed phrases are accurately translated. This innovation is expected to significantly enhance digital content accessibility across India, benefiting social media platforms, e-commerce, and more.
One of the standout features of Mayura is its approach to gender-specific language, an area where traditional models have struggled. The model includes a “gender toggle” for first-person translations, allowing users to choose appropriate gender representation, making it ideal for AI-powered voice chatbots and human agents.
Mayura’s capabilities extend beyond casual communication. It is designed for domain-specific translations in areas such as legal, scientific, and technical documents, ensuring complex jargon is accurately translated while maintaining formatting elements. This makes it particularly useful for educational and government communications.
The model is now available as an API, enabling developers and businesses to integrate it into their platforms. Mayura is part of Sarvam AI’s broader suite of products, which includes Sarvam Agents, Sarvam 2B, and Shuka 1.0.
Edited by Harshajit Sarmah
ALSO READ:
Shailaja Pathak
