BreakFrames

ComfyUI cannot load a video out of the box. There's a LoadImage node, sure, but a whole mp4? Nothing. BreakFrames is the missing import step: you hand it a file path, and it hands back every frame as a single IMAGE batch that plugs straight into anything that takes an image. It's the node you reach for when you want to do video-to-video, style-transfer an existing clip, or pull stills out of footage to feed something like AnimateDiff or a Wan first/last-frame setup.

Mechanically it's simple and that's the point. The one input is file_input, a plain string with the full path to your animation - the default C:/Videos/video.mp4 tells you the author's a Windows user, and it's worth copying the forward slashes. OpenCV's VideoCapture reads every frame it can get, flips each from BGR to RGB (OpenCV's default order is backwards for everything else in ComfyUI), converts each frame to a tensor, and concatenates them all into one batch. The single output, Frames, is an IMAGE - wire it into a KSampler for img2img, into GetKeyFrames to thin the batch down, or into a preview node to see what you got. If the path doesn't exist the node raises a FileNotFoundError in your console; if it exists but OpenCV can't decode it, you get a printed error and a None.

Install is a two-liner:

cd ComfyUI/custom_nodes
git clone https://github.com/LonicaMewinsky/ComfyUI-MakeFrame

Then restart ComfyUI. The pack's only dependency is opencv-python, and its __init__.py tries to pip-install that automatically on first import if it's missing. No model files to download - this is pure utility, from the same author as gif2gif, the old A1111 extension that did the same trick for a different UI. ComfyUI Manager also finds it if you search "ComfyUI-MakeFrame" (the pack's short title is ComfyBreakAnim).

Where people get burned: batch size. There's no frame-limit toggle here. A 30 fps video of a few minutes becomes tens of thousands of frames, and every one of them is a full-res tensor sitting in VRAM. An 18,000-frame batch will OOM a 24 GB card without ceremony. Feed this node short clips, or a GIF, and let GetKeyFrames do the triage afterward - that's the workflow the pack was built around.

One more honest caveat: this pack is a single commit from November 2023 and hasn't been touched since. It's a small personal tool, not a maintained project, and it shows in the rough edges. But it's also only a hundred lines of straightforward OpenCV, which is exactly why it still works today.