LLS Resolution Selector

Resolution selectors are the kind of node that looks like a waste of a dropdown until you've rebuilt a txt2img workflow with the resolution hardcoded in three different nodes. LLS Resolution Selector gives you one dropdown and two outputs - width and height - that you wire into whatever needs them. Change the dropdown, and every downstream node that takes its size from here follows.

The preset list covers the usual suspects: 512x512 and 768x768 squares, 1024x1024, portrait and landscape variants (512x768, 768x512, 768x1024, 1024x768), the 16:9-ish 1024x576 / 576x1024 pair, and 720p/1080p video-style frames up to 2048x2048. If none fit, set preset to Custom and type custom_width / custom_height directly - those fields step by 8, which is a nice nod to keeping dimensions latent-friendly, though you can still type any value in the 64–8192 range.

Where it plugs in

Wire width and height into the width/height inputs of a latent node, a KSampler-adjacent sizing node, or LLS Generation Config if you want family-aware defaults instead. It pairs naturally with LLS Simple Empty Latent, which has its own size presets - the selector is for when you want the resolution to be a single source of truth that several nodes read from, rather than typing it into each one.

A genuinely useful pattern: keep one Resolution Selector feeding both your Empty Latent and a Multiple of-aware downstream node (like LLS Concat By Target's multiple_of). One dropdown then controls the whole graph's canvas size, and you stop hunting for the three places a resolution is buried.

Installing it

It's part of LLS-node:

cd ComfyUI/custom_nodes
git clone https://github.com/Gin3601/LLS-node
# restart ComfyUI

Or ComfyUI Manager → search "LLS-node". Pure Python, no model downloads, nothing to configure besides the dropdown.

The honest take

This is a convenience node, not a capability node. If you're not using the LLS pack, ComfyUI-Image-Selector-style packs and even the built-in Empty Latent widgets cover the same ground. Its case is consistency inside an LLS workflow and the Custom-mode flexibility. There's one mild trap to be aware of: the selector just hands out numbers - it won't tell you whether the resolution is right for your model family, so for SD1.5 at 1024x1024 you're on your own. That's what LLS Generation Config is for; the selector is for when you already know the number.