Nodes/Allor Plugin/ImageContainer
ComfyUI Node

ImageContainer

A blank RGBA canvas you can size on command

By Nourepide·Created 3 years ago·Updated 2 years ago· 296
ImageContainer
    • IMAGE
    width512
    height512
    red255
    green255
    blue255
    alpha0.00

    This is the unglamorous node the rest of the Allor container family builds on, and the one you'll actually reach for when you need a canvas that isn't just "whatever the last image happened to be." ImageContainer makes a solid-color RGBA image at an exact width and height. That's it. But "exact size, on command" turns out to be the whole game when you're compositing, because the Allor ByContainer composite and draw nodes all read their canvas size from an image input.

    How it works

    Under the hood it creates a blank RGBA image with PIL (create_rgba_image) and converts it to a ComfyUI tensor. Nothing fancy - no models, no GPU math worth mentioning, no downloads. It's the equivalent of Photoshop's "New document" dialog, in node form.

    The inputs that matter:

    • width / height - canvas size in pixels, both default to 512.
    • red / green / blue - the fill color, each 0–255, default 255 (white).
    • alpha - 0–1 opacity of that fill. Default is 0, i.e. fully transparent.

    That last default is the trap. Every Allor draw node defaults to alpha 1 (opaque), but the container family defaults to 0 - easy to forget when you jump between them. If you wire this into a Preview or Save node and see nothing, or a checkerboard, you're looking at a transparent white canvas. Set alpha to 1.0 for an opaque backdrop.

    Why you'd use it

    Three reasons. First, a fixed-size backdrop: you want every output at exactly 1024×1024 regardless of what a sampler hands you, so you build the container, composite onto it, done. Second, you're feeding the container input of ImageCompositeAbsoluteByContainer, ImageCompositeRelativeByContainer, or the ImageDraw*ByContainer nodes - those take their dimensions from whatever image you pass, and this node is the cleanest way to supply one. Third, you want a solid color "background layer" to paste things onto (say, white for a product shot or a flat color for a matte).

    Installing Allor

    This node ships in the Allor Plugin (Nourepide/ComfyUI-Allor), an image-processing pack of ~90 nodes that has been around since early 2024. Install it once and every Allor node appears in the menu:

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

    Restart ComfyUI. Or skip the terminal: search "Allor Plugin" in ComfyUI Manager, hit install, restart. Manager is the more reliable route with this pack - a good chunk of the early Reddit threads about it are people whose manual clone went sideways, and reinstalling through Manager is the fix that kept coming up.

    Dependencies are light: rembg and onnx, for the pack's segmentation half. Nothing in this node's family touches a model.

    Gotchas that apply to the whole pack

    Two worth knowing up front. First, the repo was rebased (images stripped from history, size cut by ~344×). Great for disk space, occasionally painful for git pull - if auto-update starts failing, the README points at the docs' update-troubleshooting page, and the blunt fix is delete the folder and re-clone. Second, on first launch Allor writes config.json and a timestamp into its own folder and runs an update check; those [Allor] lines in your console are that, not an error.

    Categoryimage/container

    Inputs (6)

    NameTypeDefaultDescription
    widthINT512
    heightINT512
    redINT255
    greenINT255
    blueINT255
    alphaFLOAT0.00

    Outputs (1)

    NameTypeDescription
    IMAGEIMAGE