Prompt combiner

This is a string-joining node. It takes two bits of text and a separator, sticks them together, and outputs one combined string. prompt1 + separator + prompt2, out the other end. If you've ever wanted to keep a reusable "style" snippet in one place and glue it onto different subject prompts, this is the little tool that does it.

It's genuinely simple, but the reason it exists is real: in the Searge workflow, prompts are plain strings until an encoder turns them into conditioning, so combining text before encoding lets you build prompts modularly. Keep your quality tags in one node, your subject in another, join them here. The node's category is Searge/_deprecated_/Prompting, so it's from the older building-block set - but string concatenation doesn't go out of date, and this still works fine.

How it works

Three fields, one output. You type into prompt1 and prompt2, set what goes between them in separator, and the node returns the concatenation. The default separator is , - a comma and a space - which is exactly what you want for tag-style prompts, since that's how SDXL and its finetunes expect terms to be delimited. Change it if you're joining prose sentences and want a period or a space instead.

That's the entire mechanism. No CLIP, no weighting, no logic - just text assembly.

The inputs and outputs that matter

• prompt1 - the first (multiline) text block.
• separator - what to put between them; defaults to , .
• prompt2 - the second (multiline) text block.
• combined prompt (output) - the joined STRING.

Wire the output into a prompt encoder, or into another combiner if you're stacking several snippets.

Why bother

Modular prompting is the honest use case. A common pattern: one combiner joins your subject to a fixed style string; another joins that to a fixed quality/negative-avoidance block. Change the subject, keep everything else. It's a small quality-of-life win that keeps you from retyping the same trailing tags on every generation. For SDXL specifically, remember this mostly matters on the positive side - SDXL rewards shorter negatives, so don't use this to build a giant negative prompt out of habit.

How to install it

ComfyUI Manager: search SeargeSDXL, install, restart. Manual: python -m pip install opencv-python in ComfyUI's Python env first (a required dependency for the pack), then cd ComfyUI/custom_nodes && git clone https://github.com/SeargeDP/SeargeSDXL.git and restart. Windows portable has an installer script that also grabs the workflow's models.

No model download - it's pure text.

Where people get burned

• Expecting weighting or emphasis. This just concatenates. It won't add (term:1.2) weighting or blend concepts - that's a different kind of node. What you type is what you get, joined.
• Separator surprises. If your combined prompt reads oddly, check the separator. Joining two tag lists with a period, or two sentences with a comma, produces exactly the mess you'd expect.
• Type confusion. The output is a STRING, not CONDITIONING. It still needs to pass through an encoder before a sampler can use it.