Aspect Ratio (SD15)

Here's the SD 1.5 artist's cheat sheet as a node. Aspect Ratio (SD15) hands you a dropdown of thirteen 512-based resolutions - from 512x512 up to 1224x512, with every common aspect ratio in between - and outputs the matching width and height as integers. You pick "910x512 (16:9)", it gives you 910 and 512, and you wire those into your Empty Latent, Empty SD3, or resolution input. That's the whole job, and for SD1.5 workflows it removes a surprising amount of fiddling.

How it works

It's a lookup table wearing a node costume. The dropdown holds 13 presets, each a widthxheight (ratio) label:

512x512 (1:1), 512x682 (3:4), 512x768 (2:3), 512x910 (9:16), 512x952 (1:1.85), 512x1024 (1:2), 512x1224 (1:2.39), then the landscape mirrors: 682x512 (4:3), 768x512 (3:2), 910x512 (16:9), 952x512 (1.85:1), 1024x512 (2:1), and 1224x512 (2.39:1).

Pick one and the node returns two integers, width and height. Nothing is computed, no model loads, no VRAM is touched. The value of the node isn't the math - it's that every preset is already multiply-of-8 friendly and SD1.5-appropriate, so you're not typing "896" and hoping the sampler behaves. (A 512 base is the SD1.5 native sweet spot; the 1224 tall and wide extremes are the ones you should only use with an upscaler or a model fine-tuned for them.)

Why you'd reach for it

Same reason people reach for any "resolution preset" node: consistency and less typing. If you build workflows that get shared, a dropdown of labeled presets is friendlier than a raw width/height box, and it forces everyone onto resolutions that actually work. It also makes A/B testing trivial - flip between 1:1 and 16:9 by clicking, not by re-entering numbers. Since the outputs are plain integers, you can feed them into an Empty Latent, an Empty SD3, a latent upscale's dimensions, or any node that takes width/height inputs.

One honest note: this is specifically the SD1.5 set. If your model is SDXL or Flux, the same pack ships resolution presets tuned for those (the SDXL one starts at 1024, the "All" one mixes both) - grab the matching variant instead of stretching this one.

Installing it

Part of the RvTools v2 pack - no models, no downloads beyond the pack itself. ComfyUI Manager → search RvTools → install ComfyUI-RvTools_v2, restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/r-vage/ComfyUI-RvTools_v2

Restart and look for RvTools v2: Version 2.5.x in the console. Its extra deps (opencv-python, pilgram, pynvml, piexif) install automatically via Manager.

Troubleshooting

• If a workflow demands a resolution not in the list, just use a manual Empty Latent for that one case - the node is a convenience, not a constraint.
• Wiring to an Empty Latent: connect width and height to the matching sockets. If you only need one (say, you're fixing height while keeping width), you can wire just that output.
• Old v1 RvTools references fail in Manager - the original repo was deleted, which is why this v2 pack exists with renamed nodes. And the author's successor project, ComfyUI_Eclipse, supersedes this pack for new builds.