India’s AI development has been hindered by slow model output, limited compute resources, and expensive AI tools. However, with the launch of Varya, a video AI model developed by Avataar AI, the country is one step closer to bridging this gap.
Cheaper, Faster, and Culturally Aware Video Generation
Avataar’s Varya model is built on top of Wan 2.2, a publicly available video generation model released by Alibaba. However, unlike its predecessor, Varya uses distillation to compress the model’s capabilities into a leaner, faster version optimized for Avataar’s specific use cases.
The result is a model that runs in four steps rather than Wan 2.2’s 50, producing video 10 times faster and at a fraction of the cost. To put this in perspective, using an NVIDIA H200 GPU, Varya can generate a 5-second 720p clip in 45 seconds, compared to 1,230 seconds for Wan 2.2.
This significant improvement in speed and efficiency makes Varya an attractive option for developers who require high-quality video generation at scale.
Cultural Nuances and Affordability
One of the most striking aspects of Varya is its price. Avataar plans to charge ₹0.48 ($0.005) per second of video on its hosted service, which is a roughly 20x price difference compared to other models like Veo, Kling, Luma, and Runway.
This affordability makes Varya an attractive option for developers who require video generation at scale. Moreover, Avataar has used curated data to train Varya to recognize cultural nuances including food, clothing, architecture, and festivals.
This ensures that the model produces outputs that are culturally relevant and sensitive to local context.
India's AI Ambitions and the Role of Government Initiatives
The launch of Varya reflects a fundamental tradeoff in India’s AI ambitions. Industry veterans have noted that India can make its mark in AI by creating applications and a robust developer ecosystem rather than competing on foundation models.
The government’s initiative, India AI Mission, is part of this broader push to close the gap between India and global rivals. The mission provides selected startups with access to subsidized GPU compute in exchange for releasing their models publicly.
This has led to the development of models like Varya, which are specifically designed to meet the needs of Indian developers.
Availability and Future Plans
Varya will be released as an open-weight model on India’s AI Kosh portal, along with its training data. Developers can self-host or modify it for their own needs.
Avataar also plans to make the model available to its enterprise customers and is open to partnerships with video tools including Higgsfield and Adobe Firefly.
Anyone can try Varya now on its website using text prompts or reference images. This marks a significant milestone in India’s AI development, and it will be interesting to see how models like Varya shape the country’s AI landscape in the future.
Source: Original article