Key pose extraction
Stop scrubbing.
Start posing.
ANIMITOREF reads your video reference, tracks the movement, and pulls out the frames that actually matter — extremes, breakdowns, contacts and holds — as clean reference images you can pin next to your workspace.
- Extracts
- Key poses
- Removes
- Duplicates
- Outputs
- Sheets
Drop your video here
or click to browse
The pipeline
Every upload runs through the same six stages. Today it runs in-browser; the same contract will be served by a pose-detection backend.
- 01
Ingest video
Reading container, duration and dimensions
- 02
Analyze frames
Sampling the timeline into candidate frames
- 03
Detect movement
Measuring inter-frame motion energy
- 04
Identify key poses
Finding extremes, holds and breakdowns
- 05
Remove redundancy
Dropping near-duplicate poses
- 06
Return references
Encoding clean reference images
Chronological key poses
Poses arrive in order with frame numbers and timecode, ready to read like an X-sheet.
You stay the director
Select, deselect or delete any pose. ANIMITOREF suggests — you decide what makes the sheet.
One-click reference sheet
Composite your picks into a labelled contact sheet and download it as a PNG.