AD Image Resize

If you've used the Image Resize node from ComfyUI_essentials, you already know this one - the controls are the same, right down to the wording. It's the resize you reach for when core ComfyUI's plain "Upscale Image" isn't enough, because it handles aspect ratio, cropping, padding, and conditional resizing all in one node instead of a little cluster of them.

What makes it more than a resize

The two settings that carry the whole node are method and condition.

method decides what happens when your target dimensions don't match the image's ratio:

• stretch - squash it to fit exactly. Distorts. Rarely what you want.
• keep proportion - scale to fit inside the target without distorting.
• fill / crop - scale to cover, then crop the overflow. No distortion, no letterbox, but you lose edges.
• pad - scale to fit, then pad the empty space. No distortion, no cropping, but you get bars.

condition decides whether to resize at all:

• always resizes no matter what.
• downscale if bigger / upscale if smaller only act in one direction - great for a "cap this at 1024 but leave smaller images alone" step.
• if bigger area / if smaller area compare total pixels instead of a single dimension.

That conditional logic is the real value. It means you can drop this into a batch pipeline where images come in all sizes and it does the right thing to each one without you babysitting it.

The rest of the inputs

• image, width, height - the basics.
• interpolation - nearest, bilinear, bicubic, area, nearest-exact, lanczos. Lanczos is the clean choice for downscaling; nearest for pixel art you don't want smoothed.
• multiple_of - force dimensions divisible by this (set 8 to keep the VAE happy; 0 to ignore).
• reference_image (optional) - resize to match another image's dimensions instead of typing numbers.

Outputs are the resized IMAGE plus the final width and height it settled on - useful to wire onward when condition might have left the image untouched.

Using it well

For plain "make it bigger with a model," this isn't the node - a real ESRGAN upscaler adds detail this can't. This is geometry: fitting, cropping, padding, capping. Use it to prep inputs to a native resolution (keep proportion + downscale if bigger), to force a batch to a uniform size (fill / crop), or to match a control image to your latent (reference_image). Lanczos on the way down, and let condition skip the images that don't need touching.

Installing ComfyUI-Addoor

ComfyUI Manager: Install Custom Nodes → search "ComfyUI-Addoor" → install → restart. Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/Eagle-CN/ComfyUI-Addoor.git
cd ComfyUI-Addoor
pip install -r requirements.txt

Restart; find it under 🌻 Addoor / image. If you already run the essentials pack, note you'll now have two near-identical resize nodes - pick whichever you like, they do the same job.