Video Tagger

About this project
Video Tagger turns hours of unstructured footage into a reviewable record. Give it videos and a list of things to look for, and Gemini returns the timestamps where each one happens, written straight into the record. A reviewer then curates those findings on a scrubbable timeline, editing or deleting anything that is wrong, and builds a structured picture of which tagged moments appear, how often, and for how long. Selected moments can be rendered into shareable compilation reels.
The project was inspired by a concrete need, helping a family turn footage of their child's physical progress into a reviewable record, but the tool itself is general: any footage, any set of tags. Under the hood, a Next.js app pairs Clerk auth and per-user data with PlanetScale Postgres and Google Cloud Storage, Gemini handles the temporal localization, and a Cloud Run renderer produces the compilations.