Models in Motion

Tools

Short Python scripts that cut, walk, stack and melt your takes. Change them in Notepad, or have a chat write new ones.

How it works

  1. A folder with the takes of one shot.
  2. Copy into it RUN.bat (from the kit's tools) and the scripts you want (from tools\scripts).
  3. Double-click RUN.bat. Every .py in the folder runs, and INDEX opens with every video, named. Drag one .py onto RUN to run only that one.

The scripts

ScriptWhat it doesChange it at the top
pulse_cut.pyCuts between the takes on a rhythm.RHYTHM = "x..x..x."
crab_walk.pyOne take, forward and back. Only the order is invented.FORWARD = 3 BACK = 1
velvet.pyStacks the takes: sharp where they agree, mist where they disagree.MODE = "mean"
melt.pyRedraws every frame with a small diffusion model, feeding the last one back in. GPU, about a minute; close ComfyUI first.PROMPT = "..."
breathing_cutCuts more where the picture moves. Written by a chat from one sentence.MAX_CUTS_PER_SECOND

Open a script in Notepad: the settings are at the top, in CAPITALS. Change a number or a word, save, RUN again.

Download the tools (90 KB)

Make your own

  1. In Claude or ChatGPT, paste the box below.
  2. Paste a script that is close to your idea (click it above, copy all).
  3. Write your idea in one sentence, in any language.
  4. Save the answer in your folder of takes as yourname.py (Notepad: Save as type: All files).
  5. Double-click RUN. If it stops with an error, paste the error to the chat: "fix it".
Here is a Python script that edits AI video takes. Write a NEW one with the
same shape and my idea (at the end). The shape:
- One file. Only the standard library, numpy, cv2, PIL.
- It reads the .mp4 takes in its own folder: sorted(Path(__file__).parent.glob("*.mp4"))
- It writes .mp4 (cv2.VideoWriter) into a subfolder named after the script,
  and next to each video a .txt with the settings it used.
- Settings at the top in CAPITALS, one comment each.
- Short and readable. Print what it does.
My idea: