♾️🌊✨ Combo Parameterizer & Prompts

ESS Combo Parameterizer & Prompts is a bundle node for the SDXL base-plus-refiner era of workflows: it holds the base model's dimensions, the refiner's dimensions and aesthetic-score thresholds, and all four prompts (positive/negative for both the global and latent paths) in one place, then hands everything out on individual wires. One node, a full parameter set, no hunting across the graph.

It's the branded (♾️🌊✨) descendant of the pack's older parameter nodes, and it reads like exactly what it is: a time capsule from when SDXL's refiner pipeline forced you to keep base and refiner settings in sync manually. If you still run base+refiner SDXL, that's its whole value - a single dashboard for a notoriously fiddly two-stage setup. If you're on Flux or single-stage SDXL, this node is largely museum furniture.

Inputs and outputs

The inputs are the settings you'd otherwise scatter across the base and refiner samplers:

• base_width / base_height - base stage canvas, default 1024×1024.
• base_crop_w / base_crop_h - crop offsets for SDXL's latent crop conditioning (default 0). Leave them at 0 unless you're deliberately reproducing a specific crop.
• base_target_w / base_target_h - the target size SDXL's crop conditioning expects (default 1024×1024).
• refiner_width / refiner_height - the refiner stage's canvas.
• refiner_pascore / refiner_nascore - the aesthetic-score thresholds the refiner chain uses to decide which images the refiner samples (positive ~6.5, negative ~2.5 by default).

The optional prompt inputs carry their own descriptions in the node: PendlessG ("Positive TEXT_G, acts as main prompt and connects to refiner text input"), PendlessL ("Positive TEXT_L, acts as supporting prompt"), and the matching negatives NendlessG / NendlessL.

On the output side you get all fourteen values: Base Width/Height, Base Cropped Width/Height, Base Target Width/Height, Refiner Width/Height, Positive and Negative Refiner Aesthetic Scores, and the four prompts (Positive/Negative Text_G+refiner, Positive/Negative Text_L).

How to use it

Wire the dimensions into your base and refiner Empty Latent Images (and the KSampler crop targets), the aesthetic scores into whatever refiner-gating node you use, and the four prompts into the base and refiner CLIP Text Encode nodes - TEXT_G connects to the refiner text input, TEXT_L is the supporting prompt. All the fiddly cross-wiring that usually breaks SDXL workflows gets pinned to one node you can save and load.

Installing it

Part of Endless 🌊✨ Nodes by tusharbhutt:

cd ComfyUI/custom_nodes
git clone https://github.com/tusharbhutt/Endless-Nodes
# restart ComfyUI

Or ComfyUI Manager → search "Endless". No dependencies or model files.

Gotchas

This node is the most era-specific thing in the pack - it exists for SDXL base+refiner, and if your workflow doesn't use a refiner stage, most of its outputs are dead weight. Defaults are sane (1024×1024, pascore 6.5/nascore 2.5), so you can plug it in and adjust from there. And the pack caveat applies as always: single hobbyist maintainer, AI-assisted, mid-2025 rewrite - the parameterizer is a tool that worked for its author's SDXL pipeline, not a supported product.