ComfyUI Node

Frame Skipping

Chop the head off your clip — skipping the first N frames without a fight

By baicai99·Created 2 years ago·Updated about a year ago· 15
Frame Skipping
  • images
  • skipped_images
skip_initial_frames10

Sometimes the first frames of a clip are the problem, not the material. Frame Skipping is the "drop the first N frames" node from the ComfyUI-FrameSkipping pack - you feed it a batch of images and a skip count, and it returns everything after the first N. Mechanically it's images[skip_initial_frames:], a torch slice, nothing more. No models, no dependencies, no config files.

Why does this exist? Read the pack README and the origin story is right there: the author was building an AnimateDiff-era video where the first 16 frames had to be handled differently from everything after, to avoid animation overlap. AnimateDiff's native context is 16 frames - the famous wall - and the classic trick is processing the first 16 frames one way and the rest another way. This node is the "everything after" half of that trick. Frame Truncating (same pack) is the "keep only the first 16" half. If you're doing any per-section video processing - re-render the tail, keep the head, stitch back together - these two are a matching set.

The only input you actually set is skip_initial_frames (default 10). It's a plain integer: skip 16, get frames 16 onward. The output is skipped_images, and it plugs straight into anything that takes an IMAGE batch - another sampler pass, a VAE encode, a concat node on the way back out. That's the whole node, and that's fine. It's the kind of utility you forget is even in the graph until you need it.

The gotcha: if your skip count is greater than or equal to the total number of frames, the node returns an empty tensor. Silent, no error. You'll notice because whatever's downstream suddenly has zero frames to chew on, and it's easy to hit when you reuse a workflow built for a longer clip on a shorter one. Output mysteriously vanished? Count your frames, then check the skip value. (And note the contrast with its packmate: Frame Truncating returns your original images unchanged in the same situation. The two nodes fail differently, so don't assume.)

Install is the same as the rest of the pack: ComfyUI Manager → search "ComfyUI-FrameSkipping", or git clone https://github.com/baicai99/ComfyUI-FrameSkipping into custom_nodes, then restart. There's no requirements.txt and no model download - the whole pack runs on plain torch, which ComfyUI already ships. You'll find it under "baicai" in the node menu.

Categorybaicai/picture

Inputs (2)

NameTypeDefaultDescription
imagesIMAGE
skip_initial_framesINT10

Outputs (1)

NameTypeDescription
skipped_imagesIMAGE