India AI Mission launches indigenous Varya. What is it?

The launch was announced at a press event in New Delhi, attended by S. Krishnan, Secretary, Ministry of Electronics and Information Technology (MeitY), Government of India, along with senior leadership from Avataar.

Positioned as a India-first innovation, Varya has been developed to understand and generate culturally nuanced visual content across India’s diverse regions, festivals, communities, food practices, clothing styles, public spaces, and everyday life. The company said the model moves beyond a generalized global AI approach, focusing instead on India-specific contexts and real-world use cases.

Avataar was among the companies selected under the IndiaAI Mission to develop indigenous foundation AI capabilities. Access to subsidised national AI compute infrastructure played a key role in enabling the research behind Varya, underscoring the government’s push to strengthen domestic deep-tech innovation.

According to the company, Varya is built on a distillation-based architecture that reduces video generation steps from around 50 to just 4, significantly improving efficiency while maintaining output quality. Internal benchmarks indicate that the model can generate video at approximately ₹0.48 per second, making it up to 10 times more cost-efficient than several leading global video models.

The platform is built around a simple workflow—Idea → Video → Story—enabling users to generate videos from text prompts or images and extend them into multi-clip narratives. Its use cases span educational content, advertising, storytelling, and public information dissemination.

Speaking at the launch, MeitY Secretary S. Krishnan said the initiative marks a significant milestone in India’s AI journey, adding that Varya reflects the government’s commitment to fostering indigenous, research-led AI capabilities and enabling innovation at scale.

Avataar Co-founder and CEO Sravanth Aluru said Varya demonstrates that frontier-level video AI can be made significantly more efficient and accessible. He added that affordability is essential for building AI products at population scale in a country of over 1.4 billion people.

The company further stated that “affordable AI is inclusive AI,” emphasizing that India’s AI opportunity lies not only in building larger models, but also in developing efficient, context-aware systems designed for real-world scale and accessibility.

Avataar also announced plans to publish a technical report detailing Varya’s architecture, distillation approach, and performance benchmarks.

Varya’s core innovation lies in distilled video generation, a model compression technique in which a smaller “student” model learns from a larger “teacher” model, reducing computational complexity while preserving output quality. While traditional video models typically require more than 50 iterative steps, Varya achieves comparable results in just a few.

With this launch, Avataar aims to position Varya as a foundation for scalable AI-driven storytelling, with applications across education, marketing, entertainment, and digital services in India.