Selector Hub...

Selector Hub (class name Selector Hub, displayed "Selector Hub...") is the raw-numbers version of the pack's main Selector. Same twelve outputs - WIDTH, HEIGHT, BATCH_SIZE, STEPS, REFINER_STEPS, CFG, REFINER_CFG, DENOISE, SCALE, VARIATION_STR, SAMPLER_NAME, SCHEDULER - but instead of picking a resolution from a curated 44-entry dropdown, you type the width and height straight in. It's the Selector for when the curated list doesn't fit your case, or when you want a computed value driving the broadcast instead of a human picking from a menu.

That last bit is the real reason this node exists. The main Selector is built around a human choosing 3:2___1216x832 from a list. Selector Hub lets you feed WIDTH and HEIGHT from anywhere - an image node's actual dimensions, a math node, another workflow - because they're regular inputs, not a locked dropdown. Want to make a graph that automatically sizes the latent to match your input image? Wire the image's dimensions into Selector Hub, and every downstream consumer follows. The main Selector can't do that; Hub can.

How it works

Every input is optional, everything passes through to the matching output untouched. There's zero logic here - no ratio swapping, no validation, nothing computed. sampler_name and scheduler still come from ComfyUI's real SAMPLERS/SCHEDULERS lists, so those dropdowns stay honest, but width and height are free-form integers. Type 1344 and get 1344 out, no questions asked.

That makes it a pass-through hub in the literal sense: one node that gathers a bunch of settings into named outputs so you can broadcast them to many consumers. It's the same idea as the main Selector minus the tasteful preset layer.

Inputs & outputs that matter

• width / height (INT, optional) - the free-form dimensions. Defaults are 1×1, so wire real values or you'll generate 1×1 images. This is the trap.
• steps (default 20), cfg (default 1), scale (default 2), variation_str (default 0) - the pass-through knobs with sane-ish defaults.
• sampler_name / scheduler - ComfyUI's actual sampler and scheduler lists.
• All twelve outputs wire into Empty Latent Image, KSampler, upscalers, Detailers - same consumers as the main Selector.

Where you'd pick Hub over the main Selector

• You need programmatic or input-driven dimensions (image-aware sizing, aspect math from another node).
• Your resolution isn't in the curated list - unusual aspect ratios, fine-tune-specific sizes.
• You're mid-graph and just want a tidy gathering point for settings you're already computing elsewhere.

Installing it

Same as every node in this pack. ComfyUI Manager → search comfyui-selector → install → restart, or:

cd ComfyUI/custom_nodes
git clone https://github.com/exdysa/comfyui-selector

Restart and it's under Selector_Recourse. No requirements.txt, no model files, no network access.

Gotchas

• The default 1×1 will bite you. The main Selector defaults to a sensible 1024×1024; Hub defaults to 1×1 because it can't assume your intent. If your renders suddenly come out microscopic, check Hub's width/height.
• No rotation swap here - if you want portrait, type portrait dimensions yourself.
• It doesn't validate anything against your model. 1728×576 will generate on an SDXL model even if it's not a training resolution; you just might not love the result.
• Single-author, GPL-3.0, quiet since early 2025 - a small pack, but this node's "anything can feed it" nature is genuinely handy.