Rather than building a new video architecture from scratch, Avataar focused on compressing an existing one. The company started with Alibaba’s Wan 2.2 and applied distillation techniques that reduced the inference process from 50 steps to four. According to Avataar, that change enables video generation at roughly ten times the speed of the original model while targeting e-commerce-focused applications.
Pricing is central to the company’s strategy. Avataar is charging ₹0.48 ($0.005) per second of generated video through its hosted service. The company compares that with pricing from Veo, Kling, Luma, and Runway, which it says typically exceeds $0.10 per second. The gap positions Varya as a lower-cost alternative for organizations that need to generate video at scale.
The launch also highlights the early results of India’s AI infrastructure initiative. Avataar is one of 12 startups selected for the India AI Mission, a government program valued at roughly $1.2 billion that provides subsidized GPU resources in exchange for public model releases. The arrangement is intended to expand the country’s AI ecosystem by turning government-supported development into shared infrastructure.
Investors backing the effort argue that affordability remains a larger obstacle than model capability for many Indian use cases. Rajan Anandan, managing director at Peak XV, said widespread adoption will depend on making video generation economically viable for a broader range of users. “India is a video-first market. We see this across every large consumer internet product in India: video wins over text.”
That view aligns with a broader approach emerging among Indian startups and investors, who have increasingly focused on adapting and refining existing open models rather than competing directly with frontier AI labs on model scale. Distillation and local customization offer a less resource-intensive path to building products tailored to domestic demand.
Avataar said Varya’s training data was curated to better reflect Indian cultural references, including festivals, food, clothing, and architecture. The company argues that globally trained video models can produce outputs that overlook or generalize those details, creating an opportunity for systems optimized around local content.
Anandan said lower costs will be essential if video-generation tools are to reach a wide range of users and organizations. “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. Cost is the biggest unlock for AI adoption in India.”
Beyond its own platform, Avataar said it is pursuing enterprise deployments and is open to integrations with video tools such as Higgsfield and Adobe Firefly. The company has also made a public demonstration available through its website, where users can generate videos from text prompts or reference images.
The launch arrives as India continues to expand its AI ambitions. The India AI Mission represents a policy approach centered on subsidized computing infrastructure paired with mandatory public releases, with the goal of creating reusable resources for developers and businesses.
Varya’s performance gains come with trade-offs. Because the model is derived from Wan 2.2, its capabilities remain tied to the strengths and limitations of its parent system. Avataar is effectively exchanging higher-end output quality for faster generation and lower operating costs. That balance may be well suited to large-scale commercial workloads such as e-commerce video creation, while more demanding creative applications could continue to favor higher-fidelity systems.
The release provides one of the clearest examples so far of how India’s compute-subsidy model is translating into deployable AI products. By combining public infrastructure support with an open-weight release, Varya offers developers and enterprises a lower-cost video-generation option while testing whether affordability can become a competitive advantage in AI markets where price sensitivity remains a defining factor.
Image courtesy of Avataar AI.
This article was generated with AI assistance and reviewed for accuracy and quality.
