Prompt Processing
Unpacking Searge's prompt-weighting parameters
- parameters
- parameters
- base_conditioning_scale
- refiner_conditioning_scale
- style_prompt_power
- negative_style_power
Prompt Processing (the output node) unpacks the prompt-weighting values from the Searge parameters bundle - the conditioning scales and the style powers - into individual wires. It's the read-back partner to the Prompt Processing input node where you dial these in.
It's part of SeargeSDXL, the SDXL base-plus-refiner workflow pack Searge released at SDXL 1.0's launch in 2023. Like the rest of the output family it sits in Searge/_deprecated_/UI/Outputs - v3-era internal wiring that the v4.x data-stream rewrite replaced. It's not a node you place; it comes with the workflow.
How it works
The Searge workflow keeps every setting in one PARAMETERS bundle and uses output nodes to surface specific values where they're needed. This one carries the prompt-weighting slice. It reads the processed bundle and exposes each value as a typed float, then passes the bundle through so more output nodes can chain off it. These particular values feed into how the conditioning is built and balanced before it reaches the SDXL base and refiner.
The outputs that matter
style_prompt_power(FLOAT) - how strongly your style prompt influences the image. The one people most often want to raise.negative_style_power(FLOAT) - the same weighting for the style-side negative.base_conditioning_scale/refiner_conditioning_scale(bothFLOAT) - overall conditioning strength for each stage.parameters(PARAMETERS) - passed through for chaining.
There's nothing to set on this node - the only input is the bundle. It's pure unpacking, so the values here are whatever you chose in the input-side Prompt Processing node.
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, or the nodes won't load. The node needs no models itself; the workflow needs the usual SDXL base + refiner + fp16-fix VAE.
Common issues
"My style prompt does nothing" is the recurring Searge complaint, and while this node carries style_prompt_power, it doesn't set it - raise it in the input-side Prompt Processing node, and make sure the prompt style you picked (Misc Parameters) actually uses the style prompt. A few of the styles ignore it entirely, and no amount of power will help then.
Resist cranking base_conditioning_scale to force adherence - past a point it degrades the image instead of sharpening the concept. And the standard Searge rule: keep the workflow JSON and node version matched, since most phantom errors on the output nodes are version drift. Worth remembering that this whole prompt-weighting apparatus dates from the SDXL-refiner era; modern fine-tuned checkpoints handle style well enough that most of this fine-tuning is optional now.
Inputs (1)
| Name | Type | Default | Description |
|---|---|---|---|
| parameters | PARAMETERS | — |
Outputs (5)
| Name | Type | Description |
|---|---|---|
| parameters | PARAMETERS | — |
| base_conditioning_scale | FLOAT | — |
| refiner_conditioning_scale | FLOAT | — |
| style_prompt_power | FLOAT | — |
| negative_style_power | FLOAT | — |