FL Image Add To Batch

Pure plumbing, and there's no shame in that. FL_ImageAddToBatch takes two image batches and concatenates them into one longer batch. That's the entire job. You reach for it constantly once you start building anything with multiple image sources - stitching two video segments end to end, combining a set of generated frames with a set of reference frames, assembling a batch out of pieces that arrived on different wires. It's the + operator for image batches.

How it works

Give it images1 and images2 and it glues the second onto the end of the first, returning a single IMAGE batch. The one bit of intelligence built in: if the two batches don't share the same dimensions, it resizes images2 to match images1 (using bicubic or area interpolation depending on whether it's scaling up or down) so the concatenation is valid - you can't stack tensors of different sizes. So images1 is the boss; its dimensions win, and batch two conforms.

That's worth remembering, because it means order matters. If you want everything at batch two's resolution, put batch two first.

The inputs and output

• images1 - the first batch. Its dimensions are the target everything conforms to.
• images2 - the second batch, appended after the first and resized to match if needed.

Output is a single IMAGE batch containing batch one followed by batch two.

Installing it

Part of the Fill-Nodes pack. ComfyUI Manager: search ComfyUI_Fill-Nodes, install, restart. Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/filliptm/ComfyUI_Fill-Nodes

then restart. The "Machine Delusions" banner in the console at startup is just the pack's splash - ignore it.

Where people get tripped up

• images1 sets the resolution - order is not cosmetic. If your combined output came out at a size you didn't expect, check which batch you put first. Batch two is the one that gets resized to fit, every time.
• It only takes two inputs. To merge three or more batches, chain multiple FL_ImageAddToBatch nodes, or use the pack's FL_ImageBatch node, which takes a configurable number of image inputs. Don't hunt for a third socket on this one - there isn't one.
• Concatenate isn't composite. This stacks images into a batch (more frames), it does not overlay or blend them into a single image. If you wanted one image on top of another, that's an overlay node, not this.
• Resizing batch two can soften it. If the two batches are very different resolutions, the interpolation on batch two is a real resample - expect a little quality change on that half. Match your sources when you can.