source: techcrunch ai: cheaper, faster, and culturally aware, avataar’s video ai is built for india’s scale
level: business
avataar ai, a startup backed by peak xv, has released varya, a video generation model designed for india's scale and cultural context. the model is built by distilling alibaba's wan 2.2, compressing its capabilities into a leaner version that runs in four steps instead of 50. this makes varya about 10 times faster and far cheaper to operate. on an nvidia h200 gpu, it can produce a 5-second 720p clip in 45 seconds, compared to 1,230 seconds for wan 2.2.
the pricing is a key differentiator. avataar plans to charge 0.48 rupees per second of video, roughly $0.005, which is about 20 times less than models like veo, kling, luma, and runway that typically cost $0.10 or more per second. the company argues that current ai video models are too expensive for widespread use in india, a video-first market where cost is the main barrier to adoption for students, small businesses, and public services.
varya also addresses cultural representation, a common weakness in image and video models. avataar trained it on curated data to recognize indian festivals, food, clothing, and architecture, reducing stereotyped outputs. the model is released as open-weight on india's ai kosh portal, along with its training data, allowing developers to self-host or modify it. it is available for trial via text prompts or reference images, and the company is open to partnerships with tools like higgsfield and adobe firefly.
why it matters: drastically cheaper and faster video generation could make ai video tools accessible for population-scale applications in india, from education to e-commerce.
source: techcrunch ai: cheaper, faster, and culturally aware, avataar’s video ai is built for india’s scale