Loading article…
Avataar AI’s Varya AI video generator runs 10× faster and costs ₹0.48 per second, aiming to make video AI affordable for Indian users and developers.
Avataar AI, one of the twelve startups selected for India’s AI Mission, has unveiled Varya, a video‑generation model designed to understand local cultural contexts and operate at a fraction of the cost of existing solutions [1]. The company says the model can produce a 5‑second 720p clip in 45 seconds on an NVIDIA H200 GPU, dramatically faster than the baseline Wan 2.2 model.
Key takeaways
Avataar AI’s Varya leverages a technique called distillation to compress the capabilities of the publicly available Wan 2.2 model, reducing the inference steps from 50 to just four. This compression enables the model to generate a 5‑second, 720p video clip in 45 seconds on an NVIDIA H200 GPU, compared with 1,230 seconds required by the original Wan 2.2 model [1]. The speed gain translates into a significant cost reduction; Avataar plans to charge ₹0.48 per second of video on its hosted service, a price that is about one‑twentieth of the $0.10‑plus per‑second rates charged by rivals such as Veo, Kling, Luma and Runway [1].
Beyond speed and price, Varya is positioned as culturally aware. Avataar AI reports that it trained the model on curated data to recognize Indian festivals, food, clothing, architecture and other local nuances that generic models often miss [1]. The startup will make Varya an open‑weight model on India’s AI Kosh portal, allowing developers to download the model and its training data for self‑hosting or modification [1]. Avataar also intends to offer the model to enterprise customers and is open to partnerships with video‑tool providers such as Higgsfield and Adobe Firefly [1].
Varya’s launch highlights a strategic shift in India’s AI ecosystem: rather than competing on large foundation models, startups are focusing on application‑specific tools that can be widely adopted at lower cost. The India AI Mission’s provision of subsidized GPU compute to selected startups, including Avataar AI, aims to close the compute gap that has slowed model development in the country [1]. If Varya’s pricing and cultural relevance succeed, it could enable broader use of video AI across education, small‑business marketing, and public services, aligning with government goals to attract $200 billion in AI investment by 2028 and expand GPU capacity rapidly [1]. The open‑weight release further encourages a developer ecosystem that may accelerate innovation and localization of AI tools across India.
Coverage is mostly measured — 9 of 9 reports stay neutral.
Every Monday — the token unlocks, Fed dates & catalysts set to move crypto and markets this week. So you’re never blindsided.
Free · 3-min read · one-click unsubscribe
AI-assisted synthesis by the TrendWatcher Editorial Desk · sourced from 2 outlets · Jun 12, 2026 · How we report
The mission aims to foster local AI development by providing startups with subsidized GPU compute and infrastructure in exchange for releasing their models publicly.
Avataar uses model distillation to compress large, general-purpose models into smaller, task-specific versions that require significantly less compute power to run.
Yes, Varya is available to try on the company's website, and it will be released as an open-weight model on India's AI Kosh portal for developers.