VisionCart AI

dashboard

Inspiration

We have all seen something in a movie or series that we wanted to find later—a lamp, jacket, watch, or piece of furniture. By the time we pause, remember it, and search for it, the moment is usually gone. VisionCart makes the paused frame the starting point for discovery.

What it does

VisionCart turns a paused TV scene into a shoppable experience. Viewers can use the Fire TV remote to select objects in the frame, explore visually similar products, compare price and delivery, and save items to a session cart. A QR handoff lets them continue on their phone without losing the moment.

How we built it

We built VisionCart with React, TypeScript, Vite, and Tailwind CSS. The interface is designed for a living-room viewing distance, with large controls and clear remote-friendly states. The current demo uses deterministic scene and catalog data so judges can reliably experience the full flow without external credentials.

We also added remote-style keyboard controls: arrows select objects, Enter inspects, Space plays or pauses, S scans the frame, and J starts Judge Mode.

Challenges we ran into

Designing for a TV was different from designing for a laptop. We had to reduce the number of choices, make text readable from a distance, and ensure every action was clear without relying on a mouse.

We also wanted visual matches to feel understandable, so we focused on explanations based on details like shape, material, and footprint instead of showing an unexplained AI score.

Accomplishments that we're proud of

VisionCart is a complete interactive journey rather than a static mockup. A judge can watch a scene, pause it, select an object, understand the match, save a product, and continue on mobile.

We are also proud of the remote-first experience and the honest, repeatable demo mode.

What we learned

Designing for the living room taught us that fewer, clearer choices are more effective than a crowded interface. We also learned that explainability is part of the product experience, not just a technical feature.

What's next for VisionCart AI

Next, we want to connect the experience to a real vision pipeline that detects objects, retrieves similar catalog items, and explains each match. We also want to test VisionCart across different types of content and connect saved products to real availability and checkout.