India Unveils ‘Varya’ AI Video Model to Boost Indigenous AI Innovation

India has taken a major leap in artificial intelligence with the launch of a homegrown video generation model, ‘Varya’, aimed at delivering culturally relevant and cost-effective AI solutions.

India Launches Indigenous AI Video Model ‘Varya’

India has strengthened its push toward digital self-reliance with the introduction of Varya, a next-generation AI-powered video generation model. Developed by Bengaluru-based startup Avataar AI under the India AI Mission, the model represents a significant milestone in building sovereign AI capabilities tailored to local needs.

Announced by MeitY Secretary S. Krishnan, the launch highlights the government’s strategy to periodically release advanced AI models across various domains, ensuring India remains competitive in the global AI landscape.

What Is Varya AI?

Varya is a distilled AI video model designed to generate high-quality video content using text prompts or reference images. What sets it apart is its strong focus on understanding Indian cultural nuances, including festivals, food habits, clothing styles, and architecture.

Unlike many global AI models that often produce generic or inaccurate representations of Indian contexts, Varya is specifically trained to deliver culturally relevant outputs. This makes it particularly useful for Indian creators, businesses, and institutions.

Built on Advanced Yet Efficient Technology

Varya is based on Wan 2.2, an open-source video generation model developed by Alibaba. However, instead of using it in its original form, Avataar AI applied a process called model distillation, which compresses a large AI model into a smaller, faster, and more efficient version.

This optimisation allows Varya to generate videos in just four steps compared to the original model’s 50-step process. As a result, it delivers outputs up to 10 times faster while consuming significantly fewer computational resources.

Speed and Performance Advantages

One of Varya’s standout features is its impressive performance efficiency. When powered by an NVIDIA H200 GPU, the model can produce a 5-second 720p video clip in approximately 45 seconds. In contrast, the base model (Wan 2.2) takes over 1,200 seconds for similar output.

This massive improvement in speed makes Varya highly practical for real-world applications, especially in industries where time-sensitive content creation is critical.

Cost-Effective AI Video Generation

Affordability is another major advantage of Varya. The company plans to price its hosted service at approximately ₹0.48 ($0.005) per second of generated video. This is significantly lower than competing global platforms such as Veo, Kling, Luma, and Runway, which typically charge around $0.10 per second or more.

This roughly 20x cost advantage makes Varya accessible to startups, small businesses, and independent creators, democratizing AI-powered video production.

How to Use Varya AI

Try It Online

Users can access Varya through its official demo platform, where they can generate videos instantly. The tool allows users to experiment with prompts and visualize outputs in real time.

Two Input Methods

Varya supports two primary input methods:

• Text-to-video: Users can describe scenes or actions using text prompts.

• Image-to-video: Users can upload reference images to guide video generation.

The model performs particularly well when prompts include Indian cultural elements, delivering more accurate and context-aware results.

Open Access for Developers

Avataar AI plans to release Varya as an open-weight model on India’s AI Kosh platform. This will allow developers to:

• Download and run the model independently

• Fine-tune it for specific use cases

• Integrate it into custom applications

This open approach is expected to accelerate innovation across India’s developer ecosystem.

Enterprise and Industry Applications

Varya is designed to support a wide range of industries, including:

• Education and e-learning

• E-commerce and digital marketing

• Governance and citizen services

• Media and storytelling for MSMEs

The company is also exploring partnerships with global platforms such as Adobe Firefly and Higgsfield to expand its reach and capabilities.

Part of India’s $1.2 Billion AI Mission

The launch of Varya is part of the broader India AI Mission, a $1.2 billion initiative aimed at strengthening the country’s AI infrastructure and innovation ecosystem.

Under this program, selected startups receive access to subsidised GPU resources in exchange for contributing open AI models. Avataar AI is among 12 startups chosen for this initiative.

India’s Growing AI Ambitions

India is rapidly scaling its AI capabilities. The government aims to attract $200 billion in AI investments by 2028 while significantly expanding its GPU infrastructure.

Officials have also indicated plans to continuously release sovereign AI models across multiple domains, reinforcing India’s commitment to technological independence and global competitiveness.

Conclusion

The launch of Varya marks a pivotal moment in India’s AI journey. By combining affordability, efficiency, and cultural intelligence, the model has the potential to reshape how video content is created and consumed in India.

As AI adoption accelerates, initiatives like Varya not only empower businesses and developers but also position India as a key player in the global AI ecosystem.