Resolution Pixaroma

Resolution Pixaroma is a one-click resolution picker: a small node with a grid of aspect-ratio chips, and the moment you click one, it hands you the width and height as plain INT outputs. No math in your head, no googling "what's a good size for 16:9 SDXL," no mistyping 1024 as 1025 and then wondering why the VAE complained.

The on-node UI is a 3×3 chip grid - 1:1, 16:9, 9:16, 2:1, 3:2, 2:3, 4:3, 3:4, 4:5 - and each ratio expands to the eight most-popular sizes for that shape. The author's claim is that the presets cover SDXL, AI video, and social-media-ready dimensions in one place, and that's mostly true: it's a handy reference for the sizes people actually generate at, without you having to remember which ones are VAE-friendly.

The three modes

• Ratio chips - click a ratio, click a size, done. The 4:5 entry is there specifically because it's the Instagram-portrait shape, which is the tell that these presets are tuned for real posting habits rather than textbook theory.
• Custom Ratio - type any W:H (21:9, 16:10, whatever) and it auto-computes AI-friendly sizes for that shape. This is the mode for "I don't want to do this math."
• Custom Resolution - type exact dimensions. Math expressions work in the fields, so 1024+128 or 512*2 just evaluates - a small quality-of-life thing that runs through the whole pack.

Below all of that sit the Snap chips (16 / 32 / 64 / no snap), which round outputs to a multiple so VAE encoders stay happy. Keep snap on unless you have a specific reason to want odd dimensions; "model is fussy about dimensions" is a real thing, and rounding to a clean multiple avoids a whole class of obscure errors.

What you do with it

The two INT outputs - width and height - wire straight into an Empty Latent (or a Resolution node chain that feeds one). That's the primary job: pick the size you want to generate at without hunting for a calculator. It's also the value source for the pack's other size-hungry nodes: wire one or both outputs into Image Resize Pixaroma to force an image to that exact size, or into an image2video setup where your latent has to match a specific aspect. Because the state saves with your workflow, you can lock in the sizes you actually use and never touch it again.

Installing it

It's part of the single ComfyUI-Pixaroma pack - ComfyUI Manager search Pixaroma, Install, restart, or:

cd ComfyUI/custom_nodes
git clone https://github.com/pixaroma/ComfyUI-Pixaroma

No dependencies, no models. The node's UI is all frontend (a hidden serialized state string is the only Python-visible input), which is why there are no named inputs in the schema - don't go looking for a widget list that isn't there.

Gotchas

Short list. The math-expression support is handy but only on the Custom Resolution fields, so don't type 1024+128 into the ratio grid and expect it to parse. And remember the pack house rule: after an update, hard-refresh (Ctrl+Shift+R / Cmd+Shift+R) so the cached old UI doesn't make the node look broken. Otherwise it's a picker - the least dramatic, most-used node in the suite.