India’s progress in AI model development has lagged behind global leaders such as the U.S., Europe, and China. The domestic landscape sees only a handful of startups releasing models, predominantly focusing on large language or voice applications. To stimulate innovation and accelerate growth, the Indian government launched the India AI Mission, a substantial initiative valued at approximately $1.2 billion. This program aims to support selected startups by providing access to subsidized GPU compute, contingent on the public release of their models. Among the 12 startups chosen for this mission, Avataar AI has unveiled Varya, a new video model specifically engineered to comprehend local Indian contexts, including the identification of various festivals, culinary traditions, and attire.
Avataar AI, a startup backed by Peak XV and specializing in e-commerce video tools, did not develop Varya entirely from scratch. Instead, it leveraged Wan 2.2, a publicly available video generation model from Alibaba, and employed a technique known as distillation. This process effectively compressed Wan 2.2’s capabilities into a more streamlined, faster version, meticulously optimized for Avataar’s specific applications. The outcome is a highly efficient model that executes video generation in just four steps, a significant reduction from Wan 2.2’s 50 steps, leading to video production that is ten times faster and considerably more cost-effective.
To illustrate this efficiency with a concrete example: utilizing an NVIDIA H200 GPU, Varya can generate a 5-second 720p video clip in a mere 45 seconds. This stands in stark contrast to Wan 2.2, which requires 1,230 seconds for the same task.
Perhaps the most compelling aspect of Varya is its pricing strategy. Avataar AI intends to charge ₹0.48, equivalent to approximately $0.005, per second of video on its hosted service. This makes Varya remarkably more affordable than prominent competitors like Veo, Kling, Luma, and Runway, which typically charge $0.10 or more per second, representing an astonishing 20-fold price difference.
“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,” explained Rajan Anandan, managing director at Peak XV, in an interview with TechCrunch.
A common challenge with many image and video generation models is their tendency to overlook cultural nuances, often resulting in stereotypical or generic outputs, a phenomenon TechCrunch has previously reported. Avataar AI has proactively addressed this by utilizing carefully curated data to train Varya, enabling it to accurately recognize and incorporate diverse Indian cultural elements, including food, clothing, architecture, and festivals.
Varya is slated for release as an open-weight model on India’s AI Kosh portal, the Indian government’s central repository for publicly accessible AI models and datasets, complete with its training data. This will empower developers to self-host or customize the model for their unique requirements. Avataar AI also plans to extend Varya’s availability to its enterprise clients and is actively seeking partnerships with other video tools, including Higgsfield and Adobe Firefly. The model is currently accessible for public trial on Avataar AI’s website, where users can experiment with text prompts or reference images.
The launch of Varya underscores a pragmatic approach within India’s broader AI aspirations. Industry veterans frequently emphasize that India can carve out a significant niche in AI by concentrating on creating innovative applications and fostering a robust developer ecosystem, rather than solely competing on foundational model development. This pragmatic stance is largely driven by historical challenges in India’s model development, which has been slower than its global counterparts due to limitations in compute infrastructure and the availability of high-quality data.
The India AI Mission is an integral component of the government's comprehensive strategy to bridge these existing gaps. Last year, the mission selected 12 startups, including Avataar AI, to develop AI models, providing them with essential and cost-efficient compute resources. Earlier this year, IT minister Ashwini Vaishnaw articulated ambitious national goals, stating India's aim to attract $200 billion in AI investment by 2028 and to more than double its GPU capacity within the next six months.
The Editorial Staff at AIChief is a team of professional content writers with extensive experience in AI and marketing. Founded in 2025, AIChief has quickly grown into the largest free AI resource hub in the industry.