🎬 S42 CutFlow Ken Burns

Named after the documentary filmmaker who turned a slow, stately zoom into an entire visual language, the Ken Burns effect is how you make a still image feel like it has a camera. S42CF_KenBurns brings that to S42 CutFlow: an animated pan and zoom applied to your clip or still, with easing so the motion accelerates and decelerates like a real camera move instead of sliding at a constant rate.

The inputs that matter

The node works by cropping a window into each frame and resizing it back up, so it's all in the four corners:

• start_x / start_y - where the crop window starts (0–1, normalized across the frame). 0, 0 is top-left.
• start_zoom - the starting magnification. 1.0 is full view, 2.0 is a 2× zoom (a 50% crop).
• end_x / end_y - where the crop window ends up.
• end_zoom - the ending magnification (default 1.5).
• easing - the acceleration curve (ease_in_out by default). This is what separates a film look from a mechanical slideshow: the camera starts slow, moves, settles slow.
• output_frames - 0 means "keep the input length." Set a number to force a specific output duration.

So a classic push-in is start_zoom 1.0 → end_zoom 1.3 with start_x/y drifting from 0 to 0.3/0.2. The default values in the node actually demo this decently.

The still-image workflow (this is where it shines)

The tooltip has the key gotcha: for still images, use S42CF_ImageToClip first - otherwise KenBurns animates a single frame and you get one frozen frame out, which defeats the purpose. The right chain:

1. Image → S42CF_ImageToClip (say 72 frames, 3 seconds)
2. → S42CF_KenBurns (zoom 1.0 → 1.3, ease_in_out)
3. → grade, grain, export

That's the entire "make my AI still cinematic" recipe, and it's why this node pairs with ImageToClip constantly. It works on real video clips too - a subtle push-in on already-moving footage adds production value - but its superpower is animating stills.

Outputs are clip, frame_count, and info. Since it's a crop-resize operation on CPU, it's fast and VRAM-free, and the normalized coordinates mean the same settings work at any resolution.

Install

ComfyUI Manager → search "S42 CutFlow" → Install → restart. Or:

cd ComfyUI/custom_nodes/
git clone https://github.com/GeekyGhost/S42-CutFlow.git
pip install -r S42-CutFlow/requirements.txt

Restart fully; console prints [S42 CutFlow] Loaded 53 total nodes. Pure CPU/numpy - no models, no VRAM. Only pack dependency is OpenCV (opencv-python-headless in requirements.txt; README's troubleshooting says pip install opencv-python if you see "No module named 'cv2'").

Pack caveat, standing: Geeky Ghost runs S42 CutFlow as a work-in-progress being groomed for LTX Desktop. The composition core - KenBurns included - is stable; expanded nodes are where things move.