India’s AI sector has lagged behind the U.S., Europe, and China in model development, with only a handful of startups releasing large language or voice models. To bridge this gap, the Indian government launched the India AI Mission, a $1.2 billion initiative that provides selected startups with subsidized GPU compute in exchange for making their models publicly available.
Avataar AI, one of 12 startups chosen for the program, has now released Varya, a video generation model designed specifically for Indian market conditions. The Peak XV-backed startup, which creates video tools for e-commerce, built Varya to understand local cultural context including festivals, food, clothing, and architecture.
Rather than building from scratch, Avataar AI started with Alibaba’s publicly available Wan 2.2 video model and used a technique called distillation to compress its capabilities into a faster, more efficient version. The result is striking: Varya generates video in just four steps compared to Wan 2.2’s 50, producing content 10 times faster at a fraction of the cost.
The performance gains are substantial. Using an NVIDIA H200 GPU, Varya can generate a 5-second 720p video clip in 45 seconds, while Wan 2.2 takes 1,230 seconds for the same output. But the most significant differentiator is price.
Avataar AI plans to charge ₹0.48 ($0.005) per second of video on its hosted service – roughly 20 times cheaper than Western competitors like Veo, Kling, Luma, and Runway, which typically charge $0.10 or more per second. This pricing strategy directly targets India’s price-sensitive market.
“India is a video-first market. We see this across every large consumer internet product in India: video wins over text,” Peak XV managing director Rajan Anandan told TechCrunch. “Current AI video models are too expensive for population-scale use in India. If video AI is going to reach students, teachers, MSMEs, creators, enterprises, and public services, costs have to come down dramatically.”
The cultural awareness aspect addresses a known problem with existing image and video generation models, which often miss cultural nuances and produce stereotyped outputs. Avataar AI used curated data to train Varya to recognize specific Indian cultural elements, moving beyond the generic outputs that have plagued other AI models when generating content for diverse markets.
Varya will be released as an open-weight model on India’s AI Kosh portal, the government’s centralized repository for AI models and datasets. This means developers can self-host or modify the model for their specific needs. The company also plans to offer it to enterprise customers and is open to partnerships with video tools including Higgsfield and Adobe Firefly. Users can currently test the model on Avataar’s website using text prompts or reference images.
The launch highlights the strategic approach India is taking toward AI development. Rather than competing directly with tech giants on foundation models, industry veterans suggest India can make its mark through applications and developer ecosystem building. This pragmatic approach stems from real constraints: model development in India has been slower due to limited compute resources and quality data availability.
The India AI Mission represents the government’s broader push to address these limitations. Last year’s selection of 12 startups for subsidized compute access is part of a larger strategy. IT Minister Ashwini Vaishnaw has set ambitious targets, aiming to attract $200 billion in AI investment by 2028 and more than double India’s GPU capacity within six months.
For global AI companies, Varya’s pricing model could signal a new competitive dynamic. If Indian startups can deliver comparable quality at significantly lower costs, it may force international players to reconsider their pricing strategies for emerging markets. The success or failure of models like Varya could determine whether India establishes itself as a serious player in the global AI landscape or remains primarily a consumer of Western technology.
