With Avatar AI releasing Varya, India has faster and cheaper video AI. This model targets local festivals and commerce while providing public weight.
India faces a lag in AI model generation compared to the US, Europe, and China. Only a few startups release solutions, most of which are large-scale language or voice models. To boost development, the government has launched an approximately $1.2 billion India AI Mission. This provides selected startups with access to subsidized GPU computing power in exchange for making their models publicly available. Avatar AI, one of the 12 startups, has released Varya, a new video model designed with local context in mind, including festival awareness, cuisine, and attire.
Although Peak XV supports startups focused on creating video tools for e-commerce, we didn’t build Varya from scratch. It started by applying Wan 2.2, Alibaba’s publicly available video generation model, and Distillation, a technique that compresses the model’s functionality into a lighter, faster version and optimizes it for Avataar’s specific use case. The result is a model that runs in 4 steps instead of 50 in Wan 2.2, allowing you to create videos 10 times faster and at a significantly lower cost.
Specifically, with the NVIDIA H200 GPU, Varya can generate a 5-second 720p clip in 45 seconds, compared to 1230 seconds on Wan 2.2.
Avatar AI’s innovative approach: Varya for India
“India is a video-focused market. This is seen in all major consumer internet products in India. Video trumps text. Current AI models for video are too expensive to scale to India’s population. For video AI to reach students, teachers, small businesses, creators, enterprises, and public services, costs must come down significantly. Cost is the biggest key to AI adoption in India.”
– Rajan Anandan
Varya will be released in the open with training data on the India AI Kosh portal, a central repository of publicly available AI models and datasets from the Government of India, allowing developers to deploy or modify it to suit their needs. Avatar also plans to make this model available to enterprise customers and be open to partnerships with video tools such as Higgsfield and Adobe Firefly. Anyone can try it out right now on our website using text prompts and reference images.
Varya’s launch reflects a fundamental compromise in India’s AI ambitions. Industry veterans point out that India can contribute to AI by building applications and a strong developer ecosystem rather than competing for core models. And there is a reason for this realism. That’s because model development in India lags behind global rivals due to a lack of computing resources and limited data quality.
The India AI Mission is also part of a broader government effort to bridge this gap. Last year, 12 startups, including Avatar AI, were selected to develop AI models and provide cost-effective computing. And earlier this year, IT Minister Ashwini Vaishnaw said India aims to attract $200 billion in AI investment by 2028 and more than double its GPU capacity within six months.
If you buy through links in our articles, we may earn a small commission. This does not affect editorial independence.
The launch of Varya therefore signals India’s pragmatic approach to AI, focusing on practical applications and building a developer ecosystem with available resources for large-scale deployment of video. In the future, India can remain competitive through affordable, situational video tools that make AI more accessible to students, teachers, small businesses, creators, and public services.
