Remove Frames From Start MXD

Video models love to burn the first few frames on garbage - warping, blank lead-in, a slow fade from noise. Almost every WAN 2.2 clip I generate has a couple of unusable frames at the head. Remove Frames From Start MXD is the dead-simple answer: slice count frames off the front of an image-batch sequence and hand back the rest.

This is a utility node, and it doesn't pretend to be more. It exists because in the pack author's own WAN 2.2 workflows, trimming the head was a repeated manual step worth one dedicated node - "make existing nodes easier to use in the day-to-day," per the pack's stated mission.

What it does

Two inputs, one output:

• frames - your image batch (the IMAGE output from a VAE decode, video loader, or sampler)
• count - how many frames to drop from the start (default 10)

Output image is just frames[count:]. That's the whole mechanism: pure tensor slicing, no resampling, no re-encoding. Feed the result into a video combiner, a preview, or the next frame-processing node and you're done.

When you'd actually use it

• Post-generation cleanup - decode your WAN clip, chop the first N frames, then encode the remainder to video so the "video" starts where the action starts.
• Frame-exact trimming - you want to remove exactly the warmup frames without touching the middle, which a slider-driven editor can't do repeatably.
• Batch-friendly - it operates on the whole batch in one call, so it slots into a pipeline where every clip is getting the same trim.

The sibling node Select Frames MXD (Frames_Select_StartEnd_MXD) does the related job of grabbing a slice from the start or end; this one is specifically about dropping from the front.

The honest limitations

There's no frame counter or time display - count is in frames, so you need to know your clip's length (or its FPS) to reason about seconds. And there's no validation: ask it to remove more frames than exist and you'll get an empty tensor, which downstream nodes may or may not handle gracefully. Keep count sane relative to your sequence.

Install

Through ComfyUI Manager: search Maxed Out, install, restart. Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/Maxed-Out-99/ComfyUI-MaxedOut

Pure Python pack, no requirements.txt, no model downloads. The node lives under MXD/images in the menu. If the only thing standing between you and a clean clip is a few ugly leading frames, this is a one-node fix that beats wiring up a mask-based frame filter.