Generation Parameters

Generation Parameters is the output node that hands the Searge workflow its bread-and-butter sampler settings: seed, dimensions, steps, CFG, sampler, and scheduler. If you're tracing where the actual "how the image is sampled" numbers enter the graph, they come out of here.

It's part of SeargeSDXL, the SDXL base-plus-refiner workflow pack Searge released at SDXL 1.0's launch in 2023. It's in Searge/_deprecated_/UI/Outputs - a v3-era internal node that the v4.x data-stream rewrite replaced - and you don't drop it on the canvas yourself; it's baked into the workflow.

How it works

Searge collects every setting into one PARAMETERS bundle, then uses output nodes to unpack pieces of it where needed. This one unpacks the core generation values. You set them elsewhere (in the input nodes); this node exposes them as the individual typed wires a KSampler-style node expects - the seed goes to the seed input, steps to steps, and so on. It also passes the parameters bundle through so you can chain more output nodes.

The values themselves are the universal Stable Diffusion knobs, so they're worth knowing regardless of Searge:

The outputs that matter

• seed (INT) - the noise seed; same seed + same settings = same image.
• steps (INT) - denoising steps. SDXL is usually happy around 25–40.
• cfg (FLOAT) - prompt adherence. Too high and images get harsh/deep-fried; SDXL likes roughly 5–8.
• sampler_name (SAMPLER_NAME) and scheduler (SCHEDULER_NAME) - the sampling algorithm and step schedule.
• image_width / image_height (INT) - dimensions; keep near SDXL's ~1024-per-side sweet spot.
• save_image (ENABLE_STATE) and save_directory (SAVE_FOLDER) - passed through to the save side.
• 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, per the README. Generation needs the pack's models (SDXL base + refiner, fp16-fix VAE); the README lists them.

Common issues

Nothing to set on this node means nothing to misconfigure here directly - if your steps or CFG are wrong, change them in the input nodes, not here. This node only reports what was set.

Watch the classics that this node's values expose: CFG too high gives you contrast-blown, over-saturated images (drop it), and generating far from ~1024x1024 gives you duplicated subjects - that's where hi-res fix comes in rather than cranking width and height directly. As always with Searge, the top cause of outright errors is a workflow JSON that doesn't match the node version, so keep them in sync. And know the context: this is deprecated plumbing from the SDXL-refiner era, a period a lot of the community has moved past.