Bengaluru-based startup Avataar has launched Varya, India’s first distilled AI video generation model, offering video creation at ₹0.50 per second, at least 10 times cheaper than existing models. The startup, backed by Peak XV and operating for 12 years, unveiled the model under the government’s IndiaAI Mission, aiming to disrupt the AI video generation market dominated by global giants like OpenAI and Google, according to inc42.com.
Avataar’s approach to reducing costs does not rely on shrinking the model’s parameter count, a common method in distillation projects. Instead, the startup focused on lowering computational requirements, which is typically the most expensive part of AI video generation. CEO Sravanth Aluru explained that while most projects compress a 70-billion-parameter model to 7 billion, Avataar’s innovation lies in significantly cutting compute costs without compromising model size, enabling affordable video generation, inc42.com reported.
The AI video generation sector is currently led by large companies with massive compute budgets running into billions of dollars. Avataar’s cost-efficient model challenges this trend by proving that success depends more on minimizing compute needs than on scaling up model size. This development positions Avataar as a notable player in India’s growing AI landscape, potentially influencing how AI-generated video content is produced and consumed domestically and internationally, per inc42.com.
Avataar’s Varya model is part of the IndiaAI Mission, a government initiative supporting AI innovation. The startup’s breakthrough in reducing video generation costs to ₹0.50 per second sets a new benchmark for affordability in the sector, with the company continuing to develop its technology to expand access to AI video creation tools across India, according to inc42.com.