Meet Varya, India's own video AI model that generates videos at just Rs 0.48 per second
Avataar AI has launched Varya, a video generation model built for Indian users. The company says it is faster, cheaper and better at handling local cultural details.
Artificial intelligence has become one of the biggest technology races in the world and India is now catching up to the U.S., Europe, and China. Avataar AI, one of the 12 startups selected under the India AI Mission, has launched Varya, a new video generation AI model that aims to make video creation faster, cheaper, and more relevant for Indian users. The model has been trained to better understand Indian cultural context, from food and clothing to festivals and architecture, areas where global AI models sometimes struggle.
Built on global technology, optimised for India
Varya is based on Wan 2.2, a publicly available video generation model released by Alibaba. Avataar AI says it compressed the capabilities of Wan 2.2 into a smaller and faster version designed for its own requirements. This process, known as model distillation, allows developers to reduce computing requirements while preserving much of the original capability.
According to the company, that optimisation produced a major speed improvement. Varya runs in four steps instead of Wan 2.2’s 50. This enables video generation that is up to 10 times faster while significantly reducing costs.
To put that into practical terms, Avataar AI says that using an NVIDIA H200 GPU, Varya can generate a five-second 720p video in around 45 seconds, compared with 1,230 seconds for Wan 2.2.
The company’s internal benchmarks further claim that Varya generates video at around Rs 0.48 per second, making it substantially more cost-efficient than several leading global video models.
The company wants AI to understand India better
For users, the bigger difference may not be speed. Avataar AI says it trained Varya using curated datasets so the model can better recognise Indian cultural nuances. That includes understanding how festivals look, identifying local clothing styles, recognising food and interpreting architectural details more naturally. The goal is to reduce situations where AI-generated content looks generic or misses cultural details.
How Varya works
The experience is designed to be simple. Users can either enter a text prompt or upload an image. The model then generates a video clip, and users can continue extending the sequence by generating additional clips.
Avataar AI also said it plans to publish a technical report explaining the model architecture, distillation methodology and benchmark results.
India’s AI ambitions are expanding
Varya is also being positioned as an open ecosystem project. The company plans to release it as an open-weight model through AI Kosh, the Indian government’s central repository for publicly available AI models and datasets. This means developers will be able to host the model themselves, modify it and build applications on top of it.
Artificial intelligence has become one of the biggest technology races in the world and India is now catching up to the U.S., Europe, and China. Avataar AI, one of the 12 startups selected under the India AI Mission, has launched Varya, a new video generation AI model that aims to make video creation faster, cheaper, and more relevant for Indian users. The model has been trained to better understand Indian cultural context, from food and clothing to festivals and architecture, areas where global AI models sometimes struggle.
Built on global technology, optimised for India
Varya is based on Wan 2.2, a publicly available video generation model released by Alibaba. Avataar AI says it compressed the capabilities of Wan 2.2 into a smaller and faster version designed for its own requirements. This process, known as model distillation, allows developers to reduce computing requirements while preserving much of the original capability.
According to the company, that optimisation produced a major speed improvement. Varya runs in four steps instead of Wan 2.2’s 50. This enables video generation that is up to 10 times faster while significantly reducing costs.
To put that into practical terms, Avataar AI says that using an NVIDIA H200 GPU, Varya can generate a five-second 720p video in around 45 seconds, compared with 1,230 seconds for Wan 2.2.
The company’s internal benchmarks further claim that Varya generates video at around Rs 0.48 per second, making it substantially more cost-efficient than several leading global video models.
The company wants AI to understand India better
For users, the bigger difference may not be speed. Avataar AI says it trained Varya using curated datasets so the model can better recognise Indian cultural nuances. That includes understanding how festivals look, identifying local clothing styles, recognising food and interpreting architectural details more naturally. The goal is to reduce situations where AI-generated content looks generic or misses cultural details.
How Varya works
The experience is designed to be simple. Users can either enter a text prompt or upload an image. The model then generates a video clip, and users can continue extending the sequence by generating additional clips.
Avataar AI also said it plans to publish a technical report explaining the model architecture, distillation methodology and benchmark results.
India’s AI ambitions are expanding
Varya is also being positioned as an open ecosystem project. The company plans to release it as an open-weight model through AI Kosh, the Indian government’s central repository for publicly available AI models and datasets. This means developers will be able to host the model themselves, modify it and build applications on top of it.
