Combine (cond)

Combine (cond) is the node to grab when you want two separately-encoded prompts to both influence a generation, as parallel blocks rather than one spliced onto the other. It's the graph equivalent of A1111's AND operator - the KB's prompt-engineering essay describes AND as running two denoising paths and combining them, which made it the classic SD 1.5 / SDXL tool for multi-subject scenes. This node does that with wiring instead of syntax.

The mechanism is the least surprising thing in this pack: it takes CONDITIONING objects from its cond_N inputs and does a Python list concatenation - cond_1 + cond_2 + cond_3. In ComfyUI's conditioning format, that's how you attach separate conditioning blocks side by side; the sampler attends to all of them. It's the same operation core ComfyUI's "Conditioning (Combine)" performs, with the pack's dynamic-slot trick on top: connect a conditioning to the last slot and another one appears.

Don't confuse it with Concat (cond)

This is the distinction that trips people up, and the two nodes sit right next to each other in the menu:

• Combine (cond) - list-concatenates entire conditioning blocks. Each input stays its own block and gets attended to. Think A AND B.
• Concat (cond) - splices from's tokens onto to's tokens in one block. Think A then B as a longer prompt, the BREAK-style move.

Roughly: Combine is for two concepts you want both present; Concat is for one prompt too long to fit. If you're not sure which you need, you probably want Combine - it's the more forgiving one, and the slots mean you can stack three or four encoded blocks and each still registers.

The inputs

cond_1 (optional) and then more cond_N slots as you connect them. One output: CONDITIONING, straight into your sampler.

Like the rest of this family it's CLIP-era tooling - the AND pattern is "dead weight" on LLM-encoded models like Flux, where multi-subject scenes are handled natively and separate denoising blocks aren't a thing. On SD 1.5/SDXL, though, it's a genuinely useful way to build a prompt from separately-encoded parts (subject, style, lighting) without one giant text box.

Installing it

ComfyUI-Prompt-Helper ships no extra dependencies and needs no model downloads. ComfyUI Manager → ComfyUI-Prompt-Helper, or:

cd ComfyUI/custom_nodes
git clone https://github.com/elypha/ComfyUI-Prompt-Helper

Restart, and it's under prompt_helper.

Verdict: it's core conditioning-combine with ergonomic slots and zero surprises. If you're already reaching for ConditioningCombine, this is the same thing with a better wiring experience.