🌄 Image Resolution Adjust

Ever needed a 1024-wide image at the right height for its aspect ratio, without doing the multiplication yourself? That's a genuine recurring ComfyUI annoyance - and ImageResolutionAdjust (🌄 Image Resolution Adjust, from ControlFlowUtils) is the node that does the arithmetic for you. You give it a source width/height, it gives you back adjusted dimensions that keep the aspect ratio sensible.

One thing to get straight first: this node works on dimensions, not pixels. It doesn't resize an actual image - it computes numbers. You feed the results into an Empty Latent Image or an upscale-by-size node. That's the point: your resolution can become a computed value that responds to your workflow's logic instead of a hardcoded widget.

The inputs that matter

• source_width / source_height - the current dimensions (defaults 512×512).
• scaling_factor - multiply dimensions directly. Factor > 1 scales width up; factor < 1 changes height instead (the author's quirk, and it matters: a factor of 2 doubles width only, not the whole image). If you want uniform scaling, set the other side to 0.1-style math yourself - or just use a target dimension instead.
• target_width / target_height - the normal way to use this node. Set one of them and the other is extrapolated to preserve aspect ratio.
• tolerance - how far the final width and height may drift from the exact ratio, in pixels. The output is forced to multiples of tolerance (default 32), which keeps dimensions latent-friendly. Bigger tolerance = looser fit, more speed-friendly numbers.

The outputs are Adj. Width and Adj. Height (both INT, ready for an Empty Latent) plus Result, a string that shows the calculation it performed - handy for a text display or a quick sanity check.

The quirks worth knowing

The scaling-factor asymmetry is the real gotcha. "Scaling factor 2" does not mean 2× in both dimensions - the description says values > 1 affect width and values < 1 affect height. It's tuned for upscaling workflows where you usually want to grow in one direction. If you find the results surprising, switch to target_width/target_height, which behave the way you'd expect.

If you set both target dimensions to the same value (or equal to source), it returns them unchanged - no calculation, since none is needed. And tolerance matters more than it looks: with default 32, a 16:9 source won't come out perfectly 16:9, it'll come out the closest multiple-of-32 approximation. That's deliberate, and it's exactly what samplers want.

For real upscaling pipelines this packs nicely with the pack's other nodes - compute a target size in a Data Monitor FORMULA, then use ImageResolutionAdjust to snap it to clean values. For a beginner, the one-line summary is: set target_width to your desired width, read Adj. Height, feed both into an Empty Latent.

Install

ImageResolutionAdjust ships in VykosX/ControlFlowUtils:

# ComfyUI Manager → Install Custom Nodes → search "ControlFlowUtils"
# or manually:
cd ComfyUI/custom_nodes
git clone https://github.com/VykosX/ControlFlowUtils

Restart, find it under 🐺 VykosX-ControlFlowUtils. No dependencies, no models. The repo even ships a sample workflow for this node, which shows the target-dimension pattern better than any explanation.