I shipped binquery 0.3.0 — local semantic search for your own video footage

I built binquery because I kept running into a simple editing problem: I could remember the shot I wanted, but not its filename. Searching meant opening clips one by one.

binquery is a free, open-source Python CLI that indexes a local folder of video clips with ffmpeg and OpenCLIP ViT-B-32. You describe a shot in one sentence and it returns a ranked JSON shortlist. It does not edit, assemble a timeline, or export a finished video.

Install:

pip install binquery

Zero-footage demo:

binquery demo --out /tmp/binquery-demo

The demo generates a synthetic test video locally, then runs split → index → doctor → query. The first run may download model weights; inference and footage processing run locally.

Current limits:

• This is shortlist search, not automatic editing.

• I have not yet quality-tested it on a large real-world editing bin.

• motion_label is currently index metadata, not a query filter.

GitHub: https://github.com/jennifferweslowski-design/binquery

PyPI: https://pypi.org/project/binquery/

60-second demo: https://youtube.com/shorts/LqloM8bcqOA

I’d especially value feedback from editors and people building local-first media tools: what is the first footage query you would try?