Nodes/ComfyUI-UtilsCollection/Adjusted Resolution
ComfyUI Node

Adjusted Resolution

One node that computes both your base and upscaled resolution

By silveroxides·Created 2 months ago·Updated about 7 hours ago· 24
Adjusted Resolution
    • Width
    • Height
    • Upscaled Width
    • Upscaled Height
    • Batch
    width1024
    height1024
    batch_size1
    scale_by1.00
    multiple16

    Resolution alignment is one of those background chores that quietly wastes more time than it deserves. Your sampler wants dimensions on a 16-pixel grid, your upscale pass wants the base size scaled by a factor and re-rounded, and if you get either rounding wrong you get a cryptic error from an Empty Latent node three nodes away. UC_AdjustedResolutionParameters is the pack's answer: type in a width, height, and scale, and it hands you both the aligned base dimensions and the aligned upscaled dimensions, plus the batch size - five INT outputs, one calculation.

    What it does

    Inputs are width, height (both default 1024, 16–16384), batch_size, scale_by (default 1.0), and multiple (default 16, the grid to round to). Outputs:

    • Width / Height - your base resolution, rounded to the nearest multiple.
    • Upscaled Width / Upscaled Height - width × scale_by (and height × scale_by), again rounded to the nearest multiple.
    • Batch - passed through unchanged.

    The neat part is what it prevents. In a classic hires-fix or upscale workflow you'd hardcode "1024×1024 → 1536×1536" and then rebuild the graph every time you change the base size. Here you change one width and both the Empty Latent and the upscale step track automatically. And because the outputs are separate INTs, you can wire them anywhere - into an Empty Latent node, into a resolution picker, into a text field - without a calculator.

    The rounding behavior is worth a beat: multiple defaults to 16, which matches what ComfyUI's latent nodes expect, but it's a free parameter (4–128 in steps of 4). If you're feeding a model that wants 64-aligned inputs, set it once and every downstream number stays aligned. One honest note: rounding is to the nearest multiple, so an odd scale_by can drift your upscaled size by up to half a multiple - usually invisible, but it's there.

    Install

    Via ComfyUI Manager (search ComfyUI-UtilsCollection) or:

    cd ComfyUI/custom_nodes
    git clone https://github.com/silveroxides/ComfyUI-UtilsCollection
    

    Restart. Deps are just opencv-python and typing-extensions.

    When you'd reach for it

    Any workflow where the base resolution and the upscaled resolution should stay in lockstep - hires-fix pipelines, detail passes, batch scripts that vary size per run. It's the kind of node you install, use for a week, and then wonder how you survived typing "1536" into three separate places by hand. It replaces nothing flashy; it just removes arithmetic from your graph.

    Categoryutils

    Inputs (5)

    NameTypeDefaultDescription
    widthINT102416–16384
    heightINT102416–16384
    batch_sizeINT11–4096
    scale_byFLOAT1.000–10Scale used for the larger output resolution.
    multipleINT164–128Round the larger resolution to this pixel multiple.

    Outputs (5)

    NameTypeDescription
    WidthINT
    HeightINT
    Upscaled WidthINT
    Upscaled HeightINT
    BatchINT