Advanced Parameters

Advanced Parameters (the output side) unpacks the SDXL two-stage settings from the Searge parameters bundle - denoise, the base/refiner split, refiner strength, batch size, and the upscale factor - into individual wires the workflow can act on. It's the read-back counterpart to the Advanced Parameters input node where you set these values.

It's from SeargeSDXL, the base-and-refiner SDXL workflow pack Searge released at SDXL 1.0's launch in 2023. The Searge/_deprecated_/UI/Outputs category marks it as v3-era internal wiring; the v4.x rewrite handles this through a data stream. You never place it by hand - it's part of the workflow.

How it works

Searge bundles all settings into one PARAMETERS object and then fans specific pieces back out with output nodes. This one carries the "advanced" slice - mostly the values that control SDXL's base-then-refiner pipeline. It reads the processed bundle and exposes each value as a typed wire, and passes the bundle through for further chaining.

The values here are the ones that most shape a Searge image:

The outputs that matter

• base_ratio (FLOAT) - the base/refiner step split (0.8 = 80% base, 20% refiner). The defining SDXL two-stage knob.
• refiner_strength (FLOAT) - how hard the refiner pass acts.
• denoise (FLOAT) - full for txt2img, lowered for img2img so the source survives.
• batch_size (INT) and precondition_steps (INT) - image count and warmup noise steps.
• upscale_resolution_factor (FLOAT) and save_upscaled_image (ENABLE_STATE) - final upscale amount and whether to keep it.
• noise_offset (INT) - a noise seed offset used internally.
• parameters (PARAMETERS) - chain-through.

How to install it

ComfyUI Manager: search SeargeSDXL, install, restart. Manual: cd ComfyUI/custom_nodes && git clone https://github.com/SeargeDP/SeargeSDXL.git, then restart. Manual installs need python -m pip install opencv-python in ComfyUI's Python environment first, or the README says the pack won't load. Generation needs the models (SDXL base + refiner, fp16-fix VAE, upscalers) - the README has the links.

Common issues

There's nothing to configure on this node; it only reports values set upstream in the input node. So if base_ratio or refiner_strength is producing a worse image, change them where they're set, not here.

And that's the honest crux: the base/refiner split this node exists to carry is exactly the feature the community has largely walked away from. As one popular checkpoint author put it, the refiner "always seemed like an added step that never added much." If the refiner is smoothing your images into plastic, push base_ratio toward 1.0 or run a single fine-tuned checkpoint with no refiner at all - which is the modern default. For img2img, don't forget to drop denoise or you'll discard your source. And keep the workflow JSON and node version matched, since drift shows up as dangling outputs right on nodes like this.