India AI Mission's Varya video model generates clips at ₹0.48/sec
Synopsis
India just launched its own video-generation AI — and it costs ₹0.48 per second. Varya, built by Avataar under the IndiaAI Mission, compresses video generation from 50 steps to 4, undercutting global rivals by up to 10x on cost. It is India's clearest signal yet that the push for indigenous AI is moving from policy to product.
Key Takeaways
Varya , a distilled video-generation AI model by Avataar , was launched on 12 June under the IndiaAI Mission .
The model generates video at ₹0.48 per second , reportedly up to 10 times cheaper than leading global video models.
Varya reduces video generation from 50 steps to 4 steps using a distillation compression technique.
Development was enabled by access to subsidised national AI compute infrastructure provided by the IndiaAI Mission.
The model is designed for Indian cultural contexts — covering regions, festivals, food, clothing, and everyday life.
Avataar will publish a technical report on Varya's architecture, distillation methodology, and benchmarks.
The IndiaAI Mission on Friday, 12 June backed the launch of Varya, a distilled video-generation AI model developed by Avataar, an AI-native transformation company, with the stated aim of making 'frontier video AI affordable, accessible and relevant for India's next generation of users.' The launch took place in New Delhi in the presence of S. Krishnan, Secretary, Ministry of Electronics and Information Technology (MeitY).
What Varya Does
Varya employs a distillation technique that compresses video generation from 50 steps to just 4 steps, dramatically cutting computation time. The model generates video at ₹0.48 per second, making it reportedly up to 10 times more cost-efficient than several leading global video models, according to a statement from MeitY.
The product workflow is built around a straightforward pipeline: users type an idea, upload an image, generate a video, and extend the narrative through additional clips. As the ministry's statement described it: 'Idea → Video → Story.' A single prompt, the statement noted, can become 'a lesson, an ad, a guide, a film or a memory.'
IndiaAI Mission's Role
Avataar was among the companies selected by the IndiaAI Mission to develop indigenous foundational AI capabilities. Access to subsidised national AI compute infrastructure is credited with enabling the research that produced Varya. The initiative reflects the Centre's broader push to build homegrown AI models rather than rely entirely on foreign platforms.
'Through strategic support for foundational models, we are enabling innovation at scale and creating the building blocks for the next generation of AI solutions,' said Secretary S. Krishnan at the launch.
Cultural and Regional Design
Varya has been designed specifically for Indian contexts, built to understand and generate culturally rich visual outputs spanning India's regions, festivals, communities, food, clothing, public spaces, and everyday life. This localisation differentiates it from generic global video-generation tools that often lack cultural nuance for Indian users.
Technical Architecture and What Comes Next
Distilled video generation is a model compression technique in which a compact 'student' model replicates the outputs of a larger, slower 'teacher' model — transferring capabilities while eliminating redundant computation. Avataar has announced plans to publish a technical report detailing Varya's model architecture, distillation methodology, and benchmarks. As video increasingly becomes the dominant medium for learning, commerce, communication, and storytelling, the ability to generate high-quality content efficiently is becoming a strategic priority for Indian developers and enterprises alike.
Point of View
But the more consequential signal is structural: the IndiaAI Mission is demonstrating that subsidised compute can produce commercially competitive models, not just research prototypes. The distillation approach — cutting 50 steps to 4 — is a legitimate technical shortcut, but the real test is output quality at scale against Runway, Kling, and Sora. India has a long history of announcing indigenous tech platforms that plateau at launch; Avataar's commitment to publishing benchmarks is the right move, and those numbers will tell the real story. The cultural localisation angle is underexplored in the launch narrative — if Varya genuinely handles Indian visual contexts better than global models, that is a durable moat, not just a talking point.
NationPress
1 Aug 2026
Frequently Asked Questions
What is Varya, the AI model launched by the IndiaAI Mission?
Varya is a distilled video-generation AI model developed by Avataar and backed by the IndiaAI Mission, launched on 12 June in New Delhi. It generates video at ₹0.48 per second and is designed to be up to 10 times more cost-efficient than several leading global video models.
How does Varya's distillation technique work?
Varya uses a model compression technique where a compact 'student' model replicates the outputs of a larger, slower 'teacher' model, eliminating redundant computation. This reduces the video generation process from 50 steps to just 4 steps, significantly cutting time and cost.
What role did the IndiaAI Mission play in developing Varya?
Avataar was selected by the IndiaAI Mission to build indigenous foundational AI capabilities, and the mission provided access to subsidised national AI compute infrastructure that enabled the research behind Varya. The Centre's MeitY announced the launch officially.
Who is Varya designed for?
Varya is designed for Indian users and businesses, with the model built to generate culturally rich visual outputs across India's regions, festivals, food, clothing, and everyday life. It targets creators, educators, advertisers, and storytellers who need affordable video content.
When will Varya's technical details be made public?
Avataar has announced it will publish a technical report covering Varya's model architecture, distillation methodology, and benchmarks, though a specific release date for the report has not been confirmed.
