Image Resize by Factor

Sometimes you don't care what the final pixel dimensions are - you just want half or double what came in. That's this node. You give it an image and a multiplier, and it scales both width and height by that number. Factor 2 makes it twice as big, 0.5 makes it half, 1.5 splits the difference. No typing out "1536×1024" and doing the arithmetic yourself.

It's a small quality-of-life node, and its appeal is that it's relative. When you're chaining operations and the input size varies from run to run, "multiply by 0.5" keeps working no matter what walked in the door, whereas a fixed target size would need editing every time. Common uses: quickly downscaling a huge input before it hits a heavy step, or bumping a preview up so you can actually see it.

How it works

Straightforward multiplication. Current dimensions times the factor, resized with the interpolation mode you pick. Nothing clever happening - and that's the appeal.

Inputs and outputs

• image - the image to scale.
• factor - a float, default 1 (which does nothing). Above 1 enlarges, below 1 shrinks.
• interpolation_mode - bicubic, bilinear, nearest, or nearest exact. Bicubic is a fine default for photos; use nearest when you want crisp pixel edges (pixel art, hard-edged masks-as-images) instead of a smooth blur.

Output is a single resized IMAGE.

Installing it

It's in the comfyui_image_ops file of this pack. Grab it via ComfyUI Manager - search Various ComfyUI Nodes by Type (comfyui-various), install, restart - or clone the repo:

cd ComfyUI/custom_nodes
git clone https://github.com/jamesWalker55/comfyui-various

No models, no extra pip packages needed for this node. Restart ComfyUI once it's in place.

Common issues

The thing to keep in the back of your head is that a factor produces arbitrary dimensions. Multiply a 1017-pixel-wide image by 0.5 and you get 508, which isn't a multiple of 8. That's totally fine for a plain image, but if you're about to VAE-encode the result and sample from it, the sampler quietly prefers dimensions divisible by 8, and odd sizes can cause a one-pixel mismatch downstream. If you're feeding this into a latent, either pick factors that keep you on nice round numbers or follow it with a node that rounds to a multiple of 8.

Second, don't treat this as an upscaler. Scaling up by a factor is a dumb interpolation - it makes the image bigger, it doesn't invent detail. If you want a genuinely sharper large image, that's a job for a proper upscale model or a hires-fix pass, not a resize. This node is for cheap, honest resampling.

And the usual pack note: if the node loads red as "missing," you just don't have the pack installed yet - these JW nodes show up inside downloaded workflows all the time. Install via Manager's "Install Missing Custom Nodes" or clone the repo and restart.