Bangalore-based Avataar AI has launched Varya, one of the first homegrown Indian video AI models that generates video content at $0.005 per second, making it 27 times cheaper than comparable open-source models. Designed for scale and cultural accuracy, Varya targets India’s vast market with affordable AI tailored to local needs.
- Varya costs $0.005 per second, 27x cheaper than rivals
- Model specialized in accurately rendering Indian cultural elements
- Part of IndiaAI Mission to promote sovereign, scalable AI models
What happened
Avataar AI, a startup based in Bangalore, has released Varya, a video AI model designed to produce video content at a significantly lower cost than existing competitors. Utilizing a method called distillation, Avataar compressed Alibaba’s publicly available Wan 2.2 video generation model into a streamlined version that processes video generation in just four steps compared to 50, enabling a speed increase and substantial cost reduction.
This new model delivers video output at approximately $0.005 per second, which translates to roughly 0.48 rupees, making it 27 times cheaper than rivals such as Veo, Kling, Luma, and Runway, which typically charge $0.10 or more per second. Varya is distinct as it is trained with Indian-specific data to capture cultural details like clothing, festivals, and architecture, something global Western-trained models usually miss.
Why it matters
India’s vast market of 1.4 billion people demands AI solutions that are not only effective but also affordable and culturally relevant. Varya represents a strategic shift towards building AI technology tailored to local needs rather than relying on high-cost Western or Chinese models. This is crucial for enabling broader adoption across sectors like e-commerce, education, and public services.
By specializing in cultural specificity, Varya solves a key challenge where international models produce outputs that do not align with Indian cultural contexts, limiting their practical value locally. Additionally, being part of the government-backed IndiaAI Mission, Avataar benefits from subsidized computational resources by making its model open-weight and publicly available via India’s AIKosh repository, fostering ecosystem growth in sovereign AI development.
What to watch next
The adoption rate of Varya in Indian industries, especially in sectors reliant on localized video content, will provide insight into the viability of low-cost, culturally precise AI models as alternatives to more technically advanced but expensive foreign models. It will also demonstrate the effectiveness of India’s AI strategy focused on scalability and affordability over pure technological supremacy.
Future updates to Varya or the introduction of similar models by other startups involved in the IndiaAI Mission—such as Sarvam and BharatGen—will highlight the competitive dynamics in India’s sovereign AI space. Observers should monitor whether these models can secure significant market share and inspire additional innovation tailored to regional and cultural identities.