In a significant boost to India’s artificial intelligence ambitions, Avataar.ai has unveiled Varya, a new video generation model designed to make cutting-edge AI more affordable, faster and accessible to a wide range of users across the country.
Launched with backing from the IndiaAI Mission, Varya represents a major step forward in bringing advanced video AI into everyday use. At the event, S Krishnan, Secretary at the Ministry of Electronics and Information Technology, underlined that India is firmly on track to build its own foundational AI models, pushing back against skeptics who doubt the country’s capabilities.
He said initiatives like Varya demonstrate that India can not only participate in the global AI race but compete at the highest level. According to him, the government has already been supporting a broad spectrum of AI models, spanning beyond language into various real-world applications.
At the heart of Varya is a sophisticated technique called “distilled video generation.” In simple terms, this method allows a smaller, faster AI model to learn from a much larger and more complex one. Traditionally, video AI models require more than 50 processing steps to generate a clear output. Varya, however, reduces this to just four steps while maintaining quality. This leap dramatically cuts time, cost and computational effort, making high-end video generation far more efficient.
Avataar says this is the first time such a breakthrough has been applied to video AI at this scale in India. The company believes Varya can deliver “frontier-quality” results while being practical enough for widespread use.
What makes Varya particularly notable is its focus on India’s rich diversity. Rather than relying on a generic understanding of the country, the model has been trained to capture the nuances of different regions, cultures and lifestyles—from festivals and food to clothing and everyday scenes in towns and cities. This means users can generate videos that feel authentic and relatable across India’s varied contexts.
Sravanth Aluru, CEO and co-founder of Avataar, emphasised that the platform is built with real-world users in mind. He highlighted its potential applications across sectors—teachers creating engaging visual lessons in rural classrooms, small businesses producing advertisements, or citizens accessing government information in more accessible video formats.
Affordability is another key aspect of Varya’s promise. The company claims users can create over three minutes of video content—about 211 seconds—for every ₹100 spent, opening the door for individuals and small enterprises who previously could not afford such tools.
Technically, Varya is built on a robust architecture with 14 billion parameters, putting it among the more powerful models in its category while still maintaining efficiency through distillation. Avataar also plans to release a detailed technical report explaining the model’s design, performance benchmarks and the methodology behind its development.
As India continues to expand its AI ecosystem, Varya stands out as a sign that the country is not just adopting global innovations but actively shaping them. More importantly, it brings advanced technology closer to everyday users, turning simple ideas into vivid, shareable stories with just a few clicks.
