The challenge
Video and product discovery are often separate experiences. Connecting them calls for video processing, object detection, product data, and thoughtful playback interactions.
The workflow
Validate a video and create a local job
Prepare media and sample video frames
Return generic object classes and boxes
Reveal manually curated product links on pause
A closer look
A person uploads a video. The app samples frames and detects generic objects. A manually curated mapping connects an object class to a product entry, which appears when playback pauses.
Evidence & status
- End-to-end local video upload and processing.
- YOLO inference over sampled frames.
- Playback, thumbnails, and processing progress.
- Automated object detection stays distinct from manual product mapping.
Local proof of concept. No automatic brand recognition, dynamic ad insertion, ad-platform serving, cloud processing, or real shopping flow. The “Buy” button is illustrative only.
An engineering decision
The model detects generic object classes. A person configures each product association for each video; the model does not recognize specific brands or products.
The opportunity
The prototype explores relevant product discovery, viewer-friendly interactions, and editorial control. These are potential benefits; user research and business impact are not established.
Adapted from “NAB Oct 2026 — AI Use Cases & Stories.” Statuses and findings reflect that source. Names are listed as project contacts, not credited authors.
Open the project story on LinkedIn