Avataar AI launches Varya, a low-cost video generation model for India

In short: Avataar AI has launched Varya, a video making AI model it says can create clips much faster and at about $0.005 per second.

What happened

Avataar AI, an India-based startup that builds video tools for online shopping, launched a new AI model called Varya. A model is the engine that takes instructions and produces an output. In this case, the output is a short video clip.

The company says Varya is designed to better understand Indian context, like festivals, food, clothing, and architecture. This matters because image and video generators can produce generic or stereotyped results when they do not recognize local details.

Avataar did not build Varya from scratch. It started from Wan 2.2, a public video model released by Alibaba, then used “distillation” (like squeezing a large book into a shorter, easier-to-carry summary while keeping the key ideas) to make it smaller and faster for its needs. TechCrunch reports Varya runs in four steps instead of 50, and can generate a 5-second 720p video in about 45 seconds on an NVIDIA H200 chip, compared to 1,230 seconds for Wan 2.2.

Avataar plans to charge ₹0.48, or about $0.005, per second of video on its hosted service. That is much lower than some other popular video AI services that often charge $0.10 or more per second, according to the report.

Why it matters

If the pricing holds up in real use, it could make AI-made video affordable for more people in India, including small businesses, teachers, and creators. Avataar also says it will release Varya as an open-weight model (meaning others can download and run it themselves) through the Indian government’s AI Kosh portal.

Source: TechCrunch AI