Avataar AI, a Bangalore startup that originally built video tools for e‑commerce, announced the launch of Varya, one of the country’s first homegrown video‑generation foundation models. The new system can create video at roughly $0.005 per second, or 0.48 rupees, a price point the company says is 27 times lower than comparable open‑source offerings.
Founder Sravanth Aluru—formerly an investment banker at Deutsche Bank and an alumnus of Microsoft and IIT Mumbai—attributes the cost advantage to a process called distillation. Avataar began with Alibaba’s publicly available Wan 2.2 model, then compressed its capabilities into a leaner version that runs in four inference steps instead of the original 50. The result is a ten‑fold speed increase and a dramatic reduction in GPU usage.
Industry rivals such as Veo, Kling, Luma and Runway typically charge $0.10 or more per second for video generation. By contrast, Varya’s pricing makes high‑volume, low‑budget applications feasible for a market of 1.4 billion people where cost often outweighs top‑line performance.
Varya is not positioned to outdo U.S. or Chinese frontier models on motion realism or audio synthesis. ByteDance’s Seedance, Kuaishou’s Kling and Alibaba’s Wan push the envelope on visual fidelity, but Avataar’s pitch is scale and accessibility. The company aims to fill a gap for Indian businesses, educators and public‑sector agencies that need culturally accurate content without breaking the bank.
Cultural specificity is a core differentiator. Rather than retrofitting a Western‑trained model, Avataar curated a dataset that emphasizes Indian clothing, food, architecture, festivals and everyday settings. Global models trained primarily on Western data often produce inaccurate or culturally inappropriate outputs, limiting their usefulness in local contexts.
Varya will be released as an open‑weight model on AIKosh, the Indian government’s centralized repository for AI models and datasets. The portal is a key component of the IndiaAI Mission, a roughly $1.2 billion initiative that grants selected startups subsidized GPU compute in exchange for publicly releasing their models.
Avataar AI secured $55 million in funding from Peak XV Partners and Tiger Global. The infusion follows the company’s earlier focus on video‑editing tools for online retailers. Varya marks its first foray into foundational model development, echoing a broader trend among Indian AI startups to build sovereign technology rather than rely on foreign infrastructure.
Other Indian firms—such as Sarvam and BharatGen—have already launched their own foundational models under the same mission. The collective effort signals a strategic shift: India’s AI policy emphasizes models that serve the domestic population at affordable prices, rather than chasing the largest parameter counts.
At $0.005 per second, Varya tests whether a video model optimized for affordability and cultural relevance can achieve faster adoption than technically superior but pricier Western alternatives. Early adopters in e‑commerce, education and government are already piloting the technology, looking to generate localized video content at scale.
Analysts note that the model’s open‑weight status may encourage community contributions, potentially improving quality over time while preserving the low‑cost advantage. If Varya gains traction, it could set a new benchmark for cost‑effective AI video generation in emerging markets.
The launch underscores the growing confidence of Indian AI entrepreneurs to compete on their own terms. By marrying price efficiency with culturally aware outputs, Avataar AI hopes to demonstrate that high‑impact AI does not have to come with a high price tag.
This article was written with the assistance of AI.
News Factory APP - agentic news to boost your SEO & AEO.