Nodes/Bjornulf_custom_nodes/๐Ÿ“ฅ๐Ÿ–ผ๐Ÿ“‚ Load Images from output folder
ComfyUI Node

๐Ÿ“ฅ๐Ÿ–ผ๐Ÿ“‚ Load Images from output folder

Grab a whole folder, split by resolution

By justUmenยทCreated 2 years agoยทUpdated about a year agoยท 545
๐Ÿ“ฅ๐Ÿ–ผ๐Ÿ“‚ Load Images from output folder
    • Images resolution 1
    • Images resolution 2
    • Images resolution 3
    • Images resolution 4
    โ—„selected_folderโ–พโ–บ

    If you save your generations into named folders - one per character, one per project, whatever - this node pulls a whole folder back into the graph in one click. Pick the folder from a dropdown and it loads every image in it (non-recursively). It's the fast way to say "give me all the images I made for this character" without wiring up a dozen individual loaders.

    The quirk that makes it distinctive, and slightly hacky: it splits the images by resolution across four outputs. It's built for a specific personal workflow, and once you see why, it makes sense.

    How it works

    The node scans a folder inside your ComfyUI output directory and batches the images it finds. But ComfyUI can't stack images of different sizes into one IMAGE batch, so instead of erroring, the author sorts by resolution: all the 1024x1024s go to output one, the 512x768s to output two, and so on - up to four distinct resolutions. If a folder has fewer than four resolutions, the empty outputs return a white placeholder tensor so nothing downstream crashes. Author's own words: it's "a little hacky for now," but it lets him select all of a character's images in under a second.

    The inputs and outputs that matter

    • selected_folder (the one input) - a dropdown of the subfolders in your output directory. Pick one.

    Four outputs, all IMAGE, named by resolution bucket:

    • Images resolution 1 through Images resolution 4 - each holds the images that share one resolution. The number of outputs is fixed at four (you can't change it dynamically), which is why unused ones return white squares.

    The intended chaining trick: run these into the pack's Combine Images node if you want to merge the buckets back into one flow, or into Select an Image / Preview nodes to eyeball a set.

    How to install it

    ComfyUI Manager: search Bjornulf_custom_nodes, install, restart. Manual:

    cd ComfyUI/custom_nodes
    git clone https://github.com/justUmen/Bjornulf_custom_nodes
    pip install -r Bjornulf_custom_nodes/requirements.txt
    

    then restart. Image loading has no exotic dependency, so this node works even if the pack's optional deps don't install.

    Common issues & troubleshooting

    I'm getting white squares on some outputs. That's expected - it means the folder had fewer than four distinct resolutions, so the extra outputs return placeholder tensors instead of crashing. Only wire up the outputs you actually need.

    My folder isn't in the dropdown. The node only looks at subfolders of the ComfyUI output directory, and not recursively. Nested subfolders won't be traversed - flatten your structure or point at the direct parent.

    Images with the same look landed on different outputs. They're grouped by exact pixel dimensions, not by content. A 1024x1024 and a 1024x1536 from the same character split apart. If you want them together, feed the outputs into Combine Images.

    The dropdown didn't refresh after I saved new images. Newer versions added a refresh button for exactly this; click it, or reload the graph, so the folder list re-scans.

    More than four resolutions in one folder? Then something's overflowing - the author's own take is that if you've got more than four, you should probably be using more than one folder. Split them.

    CategoryBjornulf

    Inputs (1)

    NameTypeDefaultDescription
    selected_folderCOMBO0 options:

    Outputs (4)

    NameTypeDescription
    Images resolution 1IMAGEโ€”
    Images resolution 2IMAGEโ€”
    Images resolution 3IMAGEโ€”
    Images resolution 4IMAGEโ€”