A measured account of how a reference video moves: cuts, a timing curve, motion regions and a plain-language reading.
About
Reads a reference video or GIF (MP4, MOV, WebM, M4V, MKV or GIF) and measures how it moves. The first call reports the clip's length, frame rate, size and where the cuts are, then analyses the opening window. Each window (5 seconds by default, up to 10) comes back with a motion-over-time curve, where in the frame the motion happens, and one vision reading of a filmstrip that names what moves, when, in which direction, with what acceleration, and any logos to leave out.
It is for rebuilding the feel of an animation in your own brand. It describes motion in plain time-based terms rather than in any animation tool's vocabulary, and it stops after each window so nobody pays to analyse more than they need.
Each moving window is one paid call to a vision model; a window that holds still skips it. Extracting frames needs FFmpeg.
- TypeScript
- FFmpeg
- AI vision models
Requirements
- Runs with
- Node.js 20 or later
- Operating systems
- windows, macos, linux
- Needs installed
- FFmpeg
- Network
- Sends a filmstrip of each moving window to the vision provider over the internet.
Costs
- External costs
- OpenAI: billed to your own key — Planned direct lane: you would bring your own OpenAI key. Today the call runs through the Syvon provider proxy and Syvon credits.