h4 - Multi Image Upload

ComfyUI's stock Load Image is fine for one reference. The moment you want five - a character sheet, style refs, inpainting regions - you either stack five loaders and chase the spaghetti, or you hunt for a pack that does multi-load sanely. The h4 - Multi Image Upload is the saner option: up to ten image slots in one node, each with its own IMAGE and MASK output, and upload fields right on the widget.

It's a stripped-down slice of h4_Live's bigger all-in-one loader, and honestly it's the kind of node that quietly becomes the backbone of a workflow the moment you start doing multi-reference or inpainting work.

How it works

Under the hood it's doing exactly what you'd hope: it lists the image files in ComfyUI's input directory, gives you a dropdown per slot (image_1 through image_10) with an upload button, and loads each selected file through ComfyUI's normal image-loading path. Empty slots are handled gracefully - you get a blank placeholder tensor and an empty mask instead of an error, so you can wire all ten outputs from day one and just not fill in some of them. That's a nice property: your graph stays stable while you swap which slots are actually populated.

Each slot yields two outputs:

• IMAGE_N - the image tensor for reference, IPAdapter, image-to-image, or whatever consumer you're feeding.
• MASK_N - the alpha/mask channel. Use it when the uploaded file has transparency, or when you're feeding an inpainting region.

The inputs and outputs that matter

Honestly there's no input configuration to speak of - the ten dropdowns are the inputs. The real decision is which slot maps to which part of your workflow. A typical pattern: slots 1–3 hold character references that all feed an IPAdapter or a face-detailing branch, slot 4 holds a base image for image-to-image, slot 5 an inpainting mask source. Because every slot always outputs IMAGE + MASK, you can build the whole graph once and treat the uploads as swappable data.

Installing it

It's part of h4_Live:

cd ComfyUI/custom_nodes
git clone https://github.com/m3rr/h4_Live
# restart ComfyUI

or ComfyUI Manager → search "h4_Live." No extra dependencies.

Where it can bite you

The dropdowns read from ComfyUI's input directory, so files you upload via the node land there - that's standard, but if you drag a file in from elsewhere it may not show up until it's actually in the right folder. Also, the "10 slots" number is fixed, not dynamic: if you need eleven, this node won't grow to fit. And the blank-slot placeholder behavior, while convenient, can mask mistakes - if you forget to select an image you'll quietly generate from a tiny blank latent rather than hitting an error. That's usually harmless (you'll see it in the output) but it's worth knowing why your result looks like a solid-color test card.