Resize Image by Factor

Every other resize node in ComfyUI asks you for a target width and height. This one asks for a multiplier. Feed it an image, type 2 or 0.5, and it scales both dimensions by that factor. That sounds trivial until you've got an upscaler outputting arbitrary sizes and you just want "make it twice as big, please."

It's part of comfy-plasma by Jordach, and it shares the pack's personality: no dependencies beyond what ComfyUI already ships, one job, no ceremony.

What it does

JDC_ResizeFactor takes an IMAGE, multiplies the width and height by factor, and hands you back a new IMAGE. The factor range is 0.01–10, so it works for shrinking previews and for doubling. The aspect ratio never surprises you - it's one scalar applied to both axes, so a 512×512 image at factor 2 becomes 1024×1024, and at 0.5 becomes 256×256. No cropping, no letterboxing, no guessing.

Mechanically it's a straight Pillow resize: the tensor gets converted to a PIL image, resized, and converted back. That means the resampler dropdown genuinely matters:

• lanczos or bicubic - your defaults for upscaling anything photographic. Smooth, sharp-ish.
• nearest - the one you want for pixel art and any place you're rescaling to keep hard edges. It's also fast.
• bilinear, box, hamming - the also-rans. Bilinear is fine for downscales; the other two are niche enough that you'll probably never touch them.

The inputs that matter

There are only three, and you set two of them:

• factor - the multiplier. Default 2.
• resampler - the filter, from the list above.

That's it. The output is a single IMAGE that you can wire into VAEEncode, another post-processing node, or SaveImage. There's no batch handling here - one image in, one image out.

How to install it

No model files, no requirements.txt, nothing heavy to download. The pack only uses Pillow, numpy, torch, and requests, all of which ComfyUI already has.

Via ComfyUI Manager: open Manager → Install Custom Nodes → search comfy-plasma → install, then restart.

Or the manual way:

cd ComfyUI/custom_nodes
git clone https://github.com/Jordach/comfy-plasma

Restart ComfyUI and it'll show up under image/postprocessing.

Where people get burned

Honestly, this one's a hard node to mess up. The one thing to remember is that a factor like 0.33 won't always give you a multiple of 8, which can be annoying downstream if you're feeding a sampler that wants clean latent dims. If that's your situation, resize to an explicit target with ComfyUI's core image scale node instead. And if you're scaling up for a final output, you're probably better off with a proper upscaler than with this node's bicubic - the factor-based node is for quick geometry changes, not quality upscaling.