Avtaar.ai is launching Varya, India's first distilled video AI model, developed under the IndiaAI Mission with support from the Ministry of Electronics and Information Technology (MeitY). The model represents a significant technological breakthrough by reducing video generation from 50 steps to just 4 steps while maintaining comparable quality to leading global models, delivering more than 10x efficiency gains. This efficiency improvement makes high-quality AI video generation affordable and accessible for population-scale applications across various sectors including education, e-commerce, and communication.
The launch event will take place in New Delhi and will be attended by Shri S. Krishnan, Secretary of MeitY, along with Avtaar.ai's leadership team. Varya applies distilled video generation technology, a model compression technique where a compact student model replicates outputs of larger teacher models while eliminating redundant computation. The name Varya, derived from Sanskrit meaning "foremost" or "most excellent," reflects the ambition to create India's most valuable contribution to global AI technology.
This development demonstrates how public-private collaboration through the IndiaAI Mission's compute infrastructure can advance India's AI capabilities, focusing on making frontier AI more affordable, accessible, and relevant for real-world business and consumer use cases across high-growth markets.