Scale To Multiple Adv. 🐲

The regular Scale To Multiple node gives you one rounding mode and one scale factor for both dimensions. That covers 90% of cases. This one is for the other 10% - the times you need width and height handled differently, like snapping a wide video frame to a 16:9 grid while keeping portrait crops on their own multiple. It's the same math, with the training wheels off.

What it does

Everything from the base node carries over: scale an image (and/or integers) to a specified multiple, with crop and resize modes, optional width/height overrides, and the image-less "just give me the numbers" mode. What's added is split control:

• rounding_mode_width and rounding_mode_height - each dimension gets its own nearest/floor/ceil choice. Floor the height to cap VRAM while nearest-snapping the width, that kind of thing.
• scale_factor_width and scale_factor_height - independent pre-multipliers. If you're doing a 1.5x width pass but keeping height close to native, this is the knob that expresses it.

Those replace the single rounding_mode and scale_factor from the simple version, so the actual list of inputs you'll touch is: multiple, the two rounding modes, the two scale factors, plus image/width/height when you have them.

Why you'd pick this over the simple one

Real talk: for most workflows, the simple version is enough, and that's why it's the default recommendation. You'd step up to Advanced when:

• Your aspect ratio isn't square and you want each side snapped to its own grid.
• You're targeting a video resolution where width and height have different divisibility requirements (common on LTX/Wan pipelines that want specific multiples per axis).
• You're budget-tuning - floor one dimension to keep the total latent size under a ceiling while letting the other breathe.

The crop and resize modes (stretch, center, fill, uniform, uniform fill; lanczos down to bilinear - tensor) are identical to the base node. So the choice between the two is purely "do I need two dials or one."

Outputs

Same three: scaled_image, scaled_width, scaled_height. The width/height outputs remain usable on their own, image or no image.

Installing it

It ships in the RyuuNoodles pack:

cd ComfyUI/custom_nodes
git clone https://github.com/DraconicDragon/ComfyUI-RyuuNoodles

Restart ComfyUI, or install via Manager (search "RyuuNoodles"). No models, no extra dependencies.

Gotchas

Same traps as the simple version, worth repeating once: with image, width, and height all unconnected it raises a ValueError (it needs one source of dimensions), and image-less runs produce the "No Image" placeholder - a black PNG with text, not real output, so don't save it by mistake. Also note this is the same code base, so it shares the console logging behavior; the pack settings can quiet it if it bothers you.