Skip First Image

Skip First Image does one arithmetic operation on a batch: it drops the first frame and returns the rest. No options, no sliders, no decisions - one images IMAGE in, one IMAGE out. If you've ever generated a video clip where frame zero is a duplicate of the input image, or a frame sequence where the first frame is a reference that shouldn't be in the output, this is the node you'd wire between the source and the save.

It's a trivial node, and it knows it. But batch trimming has a way of being annoyingly fiddly in ComfyUI, where you can't just index into a batch without writing a mini-graph around it. This collapses that to one wire.

When you'd use it

• Removing a duplicated or garbage first frame from a generated video batch.
• Dropping a reference/conditioning frame that leaked into the output stream.
• Pairing with the pack's Select Last Image to trim both ends of a sequence - drop the head here, keep the tail there.

The mechanism is literally images[1:] in torch. The interesting behavior is the edge cases, and there are two:

• A one-image batch returns an empty tensor (shape preserved, batch size 0), because skipping the only frame leaves nothing. That's not an error - it's an empty IMAGE, which downstream nodes may choke on or silently skip. If that surprises you mid-workflow, this is why.
• An empty batch passes through unchanged.

The gotcha worth remembering

Because a single-frame input silently becomes an empty batch rather than failing loudly, don't rely on this node to "do nothing" to a one-frame input. If your pipeline can legitimately feed it a single frame and you expect something to survive, you want Select Last Image instead, which errors clearly when there's nothing to select. Frames trimmed, not silently zeroed.

Installation

Part of Praveen's ComfyUI Tools. Install via ComfyUI Manager (search "Praveen" / "praveen-tools"), or:

cd ComfyUI/custom_nodes
git clone https://github.com/Praveenhalder/praveen-tools

Restart ComfyUI. The pack needs nothing beyond stock ComfyUI - PyTorch, PIL, NumPy - and no model files. It's a one-liner of a node, but one-liners are what unblock a graph at 2am.