879 Teams Started. 500 Shipped. Meet the Two That Won Pixels to Products.

Three weeks ago, 879 teams signed up to build something real with Cloudinary. Only 500 finished and shipped. Today, two of them stand at the top.

Snap2Shelf took first place. Team Pravoo, with Ethnic Studio, is the runner-up. Congratulations to both teams.

The numbers behind the build

Pixels to Products, the Cloudinary AI Hackathon 2026, was a fully virtual, free-to-enter hackathon for builders across India. It drew 879 registered teams and 1,613 builders, with a prize pool of ₹1,40,000. The brief was simple and hard at the same time: build a product where Cloudinary does real work, not one that just stores images.

Teams picked from three tracks. The first was AI media pipelines, the second was generative content workflows, and the third was an open track for media-savvy startups. Every team had to submit a live demo, a public GitHub repository, a README and a demo video. Out of 879 teams, 500 made it across the line. HackIndia shortlisted the top submissions, and Cloudinary's team ran the final judging.

First place: Snap2Shelf

Snap2Shelf took the top spot by solving a problem every online seller knows. A product photo taken on a phone does not look like a brand. Snap2Shelf turns that photo into a full shelf of staged scenes while keeping the real product intact, which matters because a customer should receive what they saw. The build goes beyond static images, with a demo storefront, a director's cut, video output and an inspection view that shows what happened to each image.

It also looks like a product, not a hackathon demo. The interface is polished and the media pipeline is deep, and the team clearly understood that Cloudinary was meant to be the engine of the product, not the place files are kept.

Runner-up: Pravoo

Team Pravoo's Ethnic Studio went after a market that large platforms often ignore: small Indian sellers of ethnic wear. A seller uploads one flat photo of a garment. Ethnic Studio returns a clean cutout, several generative backgrounds, color variations, a short product video and ready-to-use exports sized for marketplaces like Meesho and Amazon and for Instagram. Buyers can then browse a catalog filtered by the same structured data the pipeline created.

The engineering matched the idea. Uploads were signed, generated variants were cached so they never rerun unnecessarily, and the transformation logic was built to be tested. It is the kind of work that shows a team understood both the user and the tools.

What both winners got right

The two projects look different, but they share three things. Both started with a real person with a real problem, in both cases someone trying to sell a product online. Both used Cloudinary for the heavy lifting, from generation and transformation to search and delivery. And both finished the story, with outputs a real user could pick up and use the same day.

Thank you, Cloudinary

A big thank you to Cloudinary, and to Head Judge Jen Looper, for backing student builders with real tools, workshops and genuine belief. Hackathons work when the sponsor shows up for the participants, and Cloudinary did.

To every team that registered, and especially the 100 that shipped: the work you submitted is what this community is about. You can see every team on the hackathon page at hackindia.org/2026/pixels-to-products-cloudinary-ai-hackathon-2026/teams.

Your turn

The next arena is already open. The AI Agents Hackathon, AgentVerse 2026, supported by OpenAI, is open for registration until 15 October, with API credits for participating teams while they last. If Pixels to Products showed you what builders can do in three weeks, this is your chance to join them.