Varya: India's Ultra-Cheap AI Video Model Built for Billions
Avataar AI's new Varya model generates video 10x faster than competitors at just $0.005 per second — a game-changer for India's massive digital economy.
India's AI Video Ambitions Get a Major Boost
India has long lagged behind the United States, Europe, and China when it comes to homegrown AI model development. While a handful of startups have released large language models and voice-based tools, the broader AI output has remained modest. To accelerate progress, the Indian government launched the India AI Mission — a $1.2 billion initiative designed to jumpstart domestic AI development by offering selected startups access to subsidized GPU computing power, provided they release their models publicly.
Among the 12 startups chosen for this program is Avataar AI, a Peak XV-backed company specializing in video tools for e-commerce. The company has now unveiled its newest creation: a video generation model called Varya, engineered specifically to understand and reflect Indian cultural context — including local festivals, traditional clothing, and regional cuisine.
How Varya Was Built
Avataar AI didn't develop Varya from the ground up. Instead, the team started with Wan 2.2, an open-source video generation model released by Alibaba, and applied a technique known as model distillation — a process that compresses the intelligence and capability of a large model into a more streamlined, efficient version tailored to specific use cases.
The outcome is remarkable. Where Wan 2.2 requires 50 processing steps to generate video, Varya completes the same task in just four steps, making it roughly 10 times faster and significantly cheaper to run.
Real-World Performance Numbers
To illustrate just how significant this improvement is: running on an NVIDIA H200 GPU, Varya can produce a 5-second clip at 720p resolution in approximately 45 seconds. The same task takes Wan 2.2 over 1,200 seconds — more than 20 minutes. For businesses and developers working at scale, that difference is enormous.
Pricing That Could Redefine the Market
Perhaps the most compelling aspect of Varya is its pricing structure. Avataar AI plans to charge just ₹0.48 (approximately $0.005) per second of generated video on its hosted platform. Compare that to established competitors like Veo, Kling, Luma, and Runway, which typically charge $0.10 or more per second — and the math becomes striking: Varya is roughly 20 times cheaper.
This aggressive pricing strategy isn't accidental. It reflects a deliberate effort to make AI video tools accessible far beyond premium enterprise clients.
"India is a video-first market. We see this across every large consumer internet product in India: video wins over text. Current AI video models are too expensive for population-scale use in India. If video AI is going to reach students, teachers, MSMEs, creators, enterprises, and public services, costs have to come down dramatically. Cost is the biggest unlock for AI adoption in India," said Rajan Anandan, Managing Director at Peak XV, in a statement to TechCrunch.
Open Access and Enterprise Partnerships
In line with its commitment to public benefit, Varya will be released as an open-weight model on India's AI Kosh portal — the government's centralized hub for publicly accessible AI models and datasets. This means developers can not only use the model freely but also self-host or customize it to fit their specific requirements. Training data will also be made available alongside the model.
For commercial applications, Avataar plans to offer Varya to its enterprise customers and is actively exploring partnerships with video production platforms such as Higgsfield and Adobe Firefly. The model is already accessible to the public on Avataar's website, where users can generate video content using text prompts or reference images.
A Pragmatic Path for India's AI Strategy
Varya's launch highlights a broader strategic reality shaping India's approach to artificial intelligence. Rather than competing head-to-head with global giants on foundation model development — a race that demands enormous compute resources and high-quality training data — many industry experts argue that India's strength lies in application-layer innovation and building a strong developer ecosystem.
The India AI Mission is central to this effort. Beyond supporting startups like Avataar AI, the initiative is part of the government's wider ambition to attract $200 billion in AI investment by 2028 and significantly expand the country's GPU infrastructure. Earlier this year, India's IT Minister Ashwini Vaishnaw stated a goal to more than double the nation's GPU capacity within six months.
For a country where hundreds of millions of people consume content primarily through video — and where cost sensitivity shapes nearly every technology adoption curve — a model like Varya could prove to be exactly the kind of locally relevant, affordable innovation that changes the game.
