XY Input: Aesthetic Score

This one sweeps an SDXL-specific micro-conditioning value called aesthetic score, which the refiner uses as a hint about how "good-looking" the target image should be. It's an obscure dial that most people never touch, and this node's honest value is letting you find out whether it does anything noticeable for your setup before you bother wiring it in. It's part of the Efficiency Nodes XY Plot system.

How it works

SDXL's conditioning carries some extra metadata beyond your prompt - original size, crop coordinates, and an aesthetic score - that the model was trained to associate with image quality. The refiner stage in particular takes a positive and negative aesthetic score. In theory, a higher positive score nudges toward more polished output. In practice the effect is subtle and often swamped by everything else, which fits the broader SDXL story: the two-stage base-plus-refiner design is the part of the model the community "promptly ignored half of."

This node steps the aesthetic score across a range so you can actually see the influence, if any. The reason this is genuinely useful rather than busywork: aesthetic score is one of those parameters people copy from a tutorial and then never question, sometimes carrying a value that does nothing for their model. A quick plot tells you whether it's a real lever or dead weight in your specific pipeline, which is worth knowing before you build it into a template you'll reuse a hundred times.

• target_ascore - positive or negative, which of the two scores you're sweeping.
• batch_count (default 3) - how many values to test.
• first_ascore / last_ascore - the range (default 0 to 10). The field technically allows up to 1000, but the meaningful action is in the low single-to-double digits; SDXL's defaults sit around 6 for positive and 2.5 for negative.

The inputs and outputs

Set target_ascore, then batch_count and the first_/last_ range. Output is X or Y (type XY) into the XY Plot node. Because aesthetic score is an SDXL refiner concept, this belongs on an SDXL Efficient KSampler workflow.

Installing it

ComfyUI Manager → search Efficiency Nodes for ComfyUI, install, restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/jags111/efficiency-nodes-comfyui

Restart ComfyUI. No downloads.

Common issues

Manage your expectations first. Aesthetic score is a weak lever - don't be surprised if a full sweep produces near-identical tiles. That's not the node failing; it's the parameter being subtle. Plotting it is actually the smart move precisely because it tells you whether it's worth caring about on your model before you commit.

It's SDXL-only in spirit - the score is part of SDXL's conditioning scheme, so plotting it on a non-SDXL workflow won't do anything useful.

And the universal XY Input rule: no output until it reaches the XY Plot node and, through it, an Efficient KSampler's script input.

Pack-wide: IMPORT FAILED at startup (the recurring pip freeze non-zero exit) takes the whole pack down. Update ComfyUI, update the node, verify the environment.