Prompts

Prompts (the output version) is an unpacker. It takes the Searge workflow's processed parameters bundle and hands back the individual prompt strings - main, secondary, style, negative, negative-style - plus the source image and mask, as separate wires you can route into the encoders and samplers. If the input-side Prompts node is where you type, this is where those prompts come back out in usable form.

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 a v3-era node; the v4.x rewrite replaced this bundle-unpack design with a data stream. You don't place it by hand - it's part of the workflow's internal wiring.

How it works

The old Searge design bundles every setting into one PARAMETERS object so the graph stays readable, then uses a family of output nodes to fan that bundle back out wherever specific values are needed. This one handles the prompt side. It reads the processed bundle - note "processed," meaning the prompt-style recipe has already combined things upstream in the Parameter Processor - and exposes the resulting strings individually. Because it also passes the parameters bundle straight through, you can daisy-chain other output nodes off it.

The outputs that matter

• parameters (PARAMETERS) - passed through so you can keep chaining output nodes.
• main_prompt, secondary_prompt, style_prompt, negative_prompt, negative_style (all STRING) - the five prompts, fanned out to feed CLIP encoders.
• image (IMAGE) and mask (MASK) - the source image and mask, for img2img/inpainting paths.

There's nothing to set here - the only input is the parameters bundle. It's pure unpacking.

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 warns the pack won't load. This node needs no models of its own; the workflow it belongs to does (SDXL base + refiner, fp16-fix VAE).

Common issues

Since this outputs the already-combined prompts, "the wrong text is coming through" is never solved here - it's decided upstream by your prompt style (Misc Parameters) and prompt powers (Prompt Processing). This node faithfully reports whatever the processor produced.

The realistic failure is version drift: the exact set of outputs on these nodes changed across Searge versions, so an old workflow JSON against a new node build can leave a wire dangling or a type mismatched right here. Fix it the standard way - latest workflow JSON, latest custom nodes. And the honest framing: this is deprecated internal plumbing from a refiner-era workflow. The five-prompt split is a genuinely useful idea, but you don't need this node, or Searge, to use it today.