Find the Moments that Matter

Search, Tag, Structure

Dataset Explorer indexes your entire video corpus, makes it searchable in plain language, and allows you to structure it with tags & collections

Hosted by us, or deployed inside your network where footage never leaves

The entire corpus is indexed, not a sample. Video and audio, one index

The index nominates candidates, a vision language model watches and ranks

Your team's internal environment, or a scoped view for outside evaluation

Works with S3, GCS, Azure, and on-prem

New footage is indexed incrementally

Built for centi-million-minute video libraries

Indexing and search run entirely server side

Capabilities

Ask for what you need in a sentence. "Hands manipulating a door handle in low light" returns matches
Results land at the clip and frame level with timecode, so you can verify what matched before exporting

Turn any result set into a named collection you can review, refine, and share across the team
Check coverage and quality across the set, then export a licensable package with a manifest

Save a query as a standing dataset definition rather than a one-time list of results
New footage entering the corpus is matched against saved definitions, so collections stay current

Text that appears in the frame is read and indexed, from signage and labels to packaging and on-screen displays
Search a word like "Cavendish" and results return timestamped to the frames where it appears