Cheaper, faster, and culturally aware, Avataar’s video AI is built for India’s scale
Original reporting by TechCrunch
India has lagged behind global leaders like the U.S. and China in AI model development, prompting its government to launch the ambitious India AI Mission. This $1.2 billion initiative aims to supercharge domestic innovation, providing crucial compute resources to selected startups in exchange for public model releases. Among the first beneficiaries, Avataar AI, a Peak XV-backed startup specializing in e-commerce video tools, has unveiled Varya, a new video generation model poised to address a critical local need: affordable, culturally aware AI.
Varya's Breakthrough
Varya stands out not just for its capabilities but its remarkable cost-effectiveness. Built using distillation on Alibaba's open-source Wan 2.2 model, Avataar AI has created a streamlined version that generates video ten times faster and at a fraction of the cost. A 5-second 720p clip, for instance, takes just 45 seconds, a dramatic improvement over its base model. This efficiency translates into a groundbreaking price point: at roughly $0.005 per second, Varya is approximately 20 times cheaper than leading international models, making AI video generation accessible to a vast, video-first Indian market. Crucially, Varya is trained to recognize Indian cultural nuances, from festivals to food, addressing a common blind spot in global AI models. Its launch as an open-weight model further underscores the nation's strategic pivot towards practical applications and a robust developer ecosystem, rather than competing solely on foundational models.
Varya's debut represents a critical milestone for India's burgeoning AI sector and a powerful validation of the India AI Mission's pragmatic strategy. By leveraging distillation to create a hyper-efficient, culturally astute video model, Avataar AI demonstrates how targeted application development, rather than a sole focus on foundational models, can yield transformative results, especially in markets constrained by compute resources. The model's unprecedented affordability and its recognition of local nuances directly address the key barriers to AI adoption in a video-first nation like India, promising to democratize advanced video generation for a vast population. This includes enabling small businesses to create compelling marketing, empowering educators with dynamic content, and fostering a boom in localized creative expression. Its open-weight release further solidifies this democratizing potential, empowering a new generation of developers to build upon its foundation.