Avataar AI Revolutionizes Video AI in India with Varya
Le brief IA que les pros lisent chaque soir
Les 7 actus IA du jour, décryptées en 5 min. Gratuit.
Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.
Choisis ton rythme
Gratuit · Pas de spam · Désabonnement en 1 clic
Avataar AI and the India AI Mission Initiative
In India, the development of AI models has lagged behind other major powers like the United States, Europe, and China. To address this situation, the Indian government has launched the India AI Mission, an initiative with a budget of approximately $1.2 billion. This mission aims to stimulate innovation by providing selected startups with subsidized access to GPU computing. Among the 12 chosen startups, Avataar AI, backed by Peak XV, stands out with the launch of its innovative video model, Varya.
Varya: A Fast and Cost-Effective Video Model
Avataar AI developed Varya based on Alibaba's Wan 2.2 model, using a technique called distillation. This method compresses the capabilities of the original model into a lighter and faster version, optimized for Avataar's specific needs. Varya operates in just four steps, compared to 50 for Wan 2.2, allowing it to generate videos 10 times faster and at a lower cost. For instance, with an NVIDIA H200 GPU, Varya can produce a 5-second clip in 720p in 45 seconds, while Wan 2.2 requires 1,230 seconds for the same result.
A Revolutionary Cost for the Indian Market
The cost of video production with Varya is particularly attractive. Avataar plans to charge only ₹0.48 ($0.005) per second of video on its hosted service. This rate is about 20 times cheaper than competing models like Veo, Kling, Luma, and Runway, which typically charge $0.10 or more per second. This cost reduction is crucial for the adoption of AI in India, where current models are often too expensive for large-scale use.
Integrating Cultural Nuances
One of the major challenges of video AI models is their tendency to produce stereotypical or generic results, often lacking cultural nuances. Avataar AI has trained Varya with carefully selected data to recognize Indian cultural specifics, such as festivals, food, and clothing. This approach helps avoid stereotypes and better meets the needs of the local market.
Accessibility and Partnerships
Varya will be released as an open-weight model on India's AI Kosh portal, along with its training data. This means developers can host it themselves or modify it according to their needs. Avataar also plans to make the model available to its enterprise clients and is open to partnerships with video tools like Higgsfield and Adobe Firefly. Users can already test Varya on Avataar's website using text prompts or reference images.
The Future of AI in India
The launch of Varya is part of a broader strategy to position India as a major player in the field of AI. Rather than competing on foundational models, India is focusing on developing applications and a robust developer ecosystem. With the goal of attracting $200 billion in AI investments by 2028, India also hopes to double its GPU capacity in the coming months. According to Rajan Anandan, managing director of Peak XV, cost is the main lever for AI adoption in India, and Varya could play a key role in this transformation.
Brief IA — L'actualité IA en français
L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.