AI2Go Multi Image Loader Advanced

The AI2Go Multi Image Loader plus two extra outputs per file: mask_N and filename_N. Same drop-up-to-eight-images node, same downscaling controls - you're gaining a mask and a name for every image you load.

The mask is why this variant exists. Inpainting in ComfyUI wants an image and a mask, and when you're processing several images that each have their own alpha channel, the plain loader can't help - you'd be extracting masks separately for every file. Here the mask comes out of the box, with stock LoadImage semantics exactly:

• Mask = 1.0 - alpha (inverted alpha), so the transparent parts of a PNG are the masked areas.
• A file with no alpha channel yields the stock 64×64 empty mask placeholder - nothing to inpaint, same as LoadImage.
• When you downscale, the mask is resized with its image, so they never drift apart.

That last one matters more than it sounds: batch inpaint only works if every mask lines up with its image, and this node guarantees it because it resizes both from the same geometry.

filename_N is the bare filename string per file, for the same bookkeeping/conditional uses as the audio Advanced variant - label your outputs, branch on which file loaded.

Everything else is shared

downscale_mode and max_size behave identically to the basic loader: off (default) passes images through untouched; keep aspect ratio / crop to square shrink only (never upscale); stretch to square forces a square and may enlarge the shorter side. output_slots defaults to auto and can be pinned so wires survive file-list edits. And the off-toggle rule is unchanged: a switched-off row keeps its socket position but emits None on image_N, mask_N, and filename_N - safe for OPTIONAL inputs, a guaranteed failure for REQUIRED ones. count = enabled rows actually emitted.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/Little-God1983/ComfyUI-AI2Go-Utils

Restart ComfyUI, or install through Manager ("ComfyUI-AI2Go-Utils"). No extra dependencies beyond what ComfyUI ships. One thing to know before you build a workflow on it: the repo is frozen at v1.6.1 with issues closed; the author's ongoing work lives at ComfyUI-IntoTheLatent-Utils under renamed ITL* nodes. This pack still works fine for legacy graphs.

Troubleshooting

• Mask is all zeros / nothing inpaints: the source images have no alpha channel. This is faithful to stock LoadImage - add transparency to the PNGs (or don't rely on per-image alpha and mask elsewhere in the graph).
• "not found in the input folder - re-add it": the file moved or was deleted from input/. Re-drop it.
• Unexpectedly small images: downscale_mode is on - it only ever shrinks.

If your batch pipeline feeds each image plus its own mask into the same inpaint branch, this is the node to grab. The mask-follows-image guarantee is the part you'd otherwise be debugging at 2 a.m.