Simple Size Picker

SimpleSizePicker is the node for the times you've typed 1216 x 832 from memory, gotten a squashed portrait instead of a landscape, and sworn at your keyboard. It's a dropdown of 34 pre-baked resolutions that spits out clean width and height integers. That's the whole job, and it does it well enough that you'll stop hand-typing dimensions everywhere.

Why does a resolution dropdown exist at all? Because diffusion models have a native resolution, and generating far off it is where the classic failures come from - duplicated anatomy, stretched limbs, tiling. SDXL was trained on a specific set of aspect buckets (1024×1024, 1216×832, 1344×768, and friends), so picking from a curated list beats inventing ratios. That's the editorial work this node does for you: every preset is a real, model-sane dimension rather than a number you hoped would work.

The mechanism is about as simple as it gets. The node ships with a static list of resolutions (5 square, 7 portrait, 9 landscape, plus widescreen, ultrawide, super-ultrawide, and panoramic entries up to 9600×1200), and execute() splits the label on " x " to hand back two integers. There's no model involved, no latent, no state - pick 1024 x 1024 (1:1) and out come 1024 and 1024.

The one input that matters:

• resolution - a dropdown with 34 choices, grouped square → portrait → landscape → ultrawide → panoramic. Tooltips tell you the ratio. Default is 512 x 512 (1:1).

Outputs are width and height, both INT. You wire them into an EmptyLatentImage node (or any node that takes pixel dimensions - an upscaler, a conditioning image size, whatever your workflow uses). It's a pure convenience node, so the outputs are the point.

Install it the ComfyUI way: Custom Nodes Manager → search ComfyUI_BishaNodes → install → restart. Or, from your ComfyUI/custom_nodes folder:

git clone https://github.com/vladpro3/ComfyUI_BishaNodes.git

Then restart ComfyUI. No models to download, no extra pip packages - the pack's dependency list is empty and it's MIT-licensed.

Two things to know before you trust the list blindly. First, the presets lean heavily toward SDXL's trained aspect ratios, which is fine - but if you're on Flux, remember it wants dimensions divisible by 64, and a few entries here (looking at you, 1080 x 1920 and 3440 x 1440) aren't. The ≈3:4 labels are also rounded, so 832 x 1216 isn't exactly 2:3. None of that matters for SDXL; it's just worth knowing the list is "good defaults," not gospel. And second, this node gives you numbers, not a latent - if you want the same dropdown to hand you a ready-made blank latent too, that's the sibling node Empty Latent Size Picker, which covers both in one shot.