TheoreticallyMotion Control
My free pose + depth motion-control tool — turn any footage into an OpenPose skeleton and depth map for video-to-video.
What it is
TheoreticallyMotion Control is a free tool I built and run locally: drop in any footage and it bakes out an OpenPose skeleton plus a depth map, packaged as a single video you can feed as a reference into Seedance, Runway, or anything else that accepts a video input. Instead of hoping the model reads your prompt’s choreography, you hand it the actual motion.
The idea came out of a rabbit hole: re-skinning real footage by driving a generation with pose and depth instead of the source pixels. Once the experiment worked, the obvious move was to stop paying per-experiment and build an app that does the baking for free, any time.
How it works
Drag a video in and it starts tracking motion immediately. Pose comes out as an OpenPose skeleton; depth has two modes — a fast one, and a consistent mode powered by Video Depth Anything. The VDA mode needs a Chromium browser (Chrome, Edge, Arc) and pulls the model from Hugging Face on first run — about 30 seconds of setup, quicker after that.
From there:
- Multi-character tracking — up to five characters, each color-coded, so you can tell the model “character in green, character in red” and it can track who you mean.
- In/out points — render just the section you need. Fewer cuts means fewer chances for the generator to screw up.
- Smoothing, tracking, and confidence controls — crank confidence when characters occlude or pass each other and the color coding gets flickery.
What it’s for
The proof run was the hardest video-to-video test I’ve put on the channel — heavy movement, contact, characters leaving and re-entering frame. Straight video-to-video generators mostly fall apart on it; with the pose + depth reference, Seedance held on. And you’re not limited to swapping a character — the whole scene can go: same fight choreography, new fighters, derelict spaceship.
One honest caveat: Seedance occasionally ignores the video input entirely on an identical prompt. Might be a context issue; hoping 2.5 sorts it. The full walkthrough is in the video and the companion post.
Get it
The whole package is on Gumroad — put $0 in the box and it’s yours. You get the pose UI, a user guide, and all the source code, so you can modify it, extend it, or gut it and build your own. Donations welcome, never obligated. Suggestions for the next version go in the video comments.