IndiaAI Mission-backed startup Avataar has launched Varya, a low-cost AI video generation model aimed at expanding access to video creation tools in India.
Imphal, June 12: India took another step in its effort to build domestic artificial intelligence capabilities with the launch of Varya, a video story-generation model developed by Avataar under the IndiaAI Mission. The model was unveiled in New Delhi in the presence of Secretary, Ministry of Electronics and Information Technology (MeitY), S. Krishnan.
According to the ministry, Varya is a distilled AI video generation model designed to significantly reduce the computing resources required for video creation. The company claims the model can generate videos using just four inference steps compared to around 50 steps required by conventional approaches, resulting in substantially lower operating costs.
Developed by Avataar, one of the startups selected under the IndiaAI Mission's initiative to support indigenous foundation models, Varya has been trained on datasets intended to reflect India's cultural diversity, including regional traditions, festivals, food, attire and social settings.
The company said the model is capable of converting user prompts and images into short videos and stories through what it describes as an "Idea → Video → Story" workflow. Potential applications include content creation, e-commerce, education, public communication and digital marketing.
At the launch, S. Krishnan said the release of one of the foundational AI models supported under the IndiaAI Mission represented an important milestone for India's AI ecosystem. He said the initiative demonstrated how government-supported computing infrastructure could help Indian startups build advanced AI technologies.
Avataar co-founder and CEO Sravanth Aluru said affordable AI would be essential for large-scale adoption in India. He argued that efficient models could play a significant role in bringing AI-powered tools to a wider user base rather than relying solely on larger and more expensive systems.
The company claimed that Varya can generate video at a cost of approximately ?0.48 per second and is up to ten times more cost-efficient than several leading global video-generation models based on internal benchmarks. Avataar said it plans to release a technical report detailing the model architecture, distillation techniques and performance metrics.
While the launch marks progress in India's efforts to build home-grown generative AI models, early testing suggests Varya's current strengths may lie more in affordability and accessibility than in matching the quality of leading international video-generation systems.
In a basic image-to-video test involving two seated individuals, the model successfully generated a handshake when prompted to make the subjects "shake hands and stand up." However, it did not complete the second part of the instruction, producing only the handshake sequence within the five-second video.
The result points to a common challenge faced by smaller and highly compressed video-generation models. Such systems can often animate simple gestures and movements but may struggle with multi-stage actions that require more advanced motion planning and temporal consistency. Actions such as standing up involve coordinated full-body movement, changing poses and interaction with surrounding objects, making them significantly more complex than arm or facial movements.
The performance also reflects the trade-off inherent in model distillation. By reducing computational requirements and generation costs, distilled models can become more accessible for wider deployment, but this may come at the expense of visual fidelity, motion smoothness and adherence to complex prompts.
For India's AI ecosystem, however, the significance of Varya may extend beyond immediate performance comparisons with global leaders. The model represents one of the first visible outcomes of the IndiaAI Mission's effort to support indigenous foundation models and reduce dependence on foreign AI infrastructure.
As Indian AI developers compete in a field currently dominated by large international firms, the challenge will be balancing affordability with the increasingly high quality expectations set by advanced video-generation systems already available globally.
