NAI 正面提示词拼接
Snap two prompts together without the typo
- a
- b
- 正面提示词
Every serious NovelAI workflow ends up with a character block, a clothing block, a scene block, and a style block, and you will eventually get tired of retyping all four into one box. NAIPositivePromptJoinV2 is the glue that puts them back together - a boring node with a genuinely useful job.
What it is
It takes two NAI_POSITIVE_PROMPT inputs (a and b) plus a separator string (default ", ") and outputs one combined NAI_POSITIVE_PROMPT. The join happens on the prompt objects, not on raw text, which is the part that matters: it preserves the polarity type, so what comes out can only feed a positive input downstream. No accidental a, b where you meant b, a either - the order is whatever you wire it in.
Two prompts joined like this feed into another join, which is how you build a chain: base character prompt → add outfit → add scene → add style, one node per addition. The pack even ships a negative twin, so your negative prompt can be assembled with the same building blocks.
When you'd reach for it
Three situations, mostly.
- Modular prompts. Keep the character tags in one node and the environment in another. Swap a scene by editing one box instead of hunting through a long string.
- Reuse across workflows. The same
1girl, masterpiecebase node can feed several different sampler branches, and each branch adds its own suffix via a join. That's the plumbing pattern of breaking one value out so it has a single source of truth - the KB calls it the "one source, many consumers" shape, and it's exactly why ComfyUI lets you convert widgets to inputs. - Batch and XY work. The XY plot nodes let you append a prompt suffix per cell, but they append to a base prompt. If your base is itself a join of a few parts, you get a clean "fixed base + variable tail" split instead of a monster string.
The inputs that matter
Honestly, just two:
a/b- the two prompt ports. Order matters, so wire deliberately.separator-", "is the right default for tags. You can change it, but NovelAI is a comma-tag model, so don't bother unless you know why.
Output is one NAI_POSITIVE_PROMPT plus nothing else - this node is terminal plumbing, not a debug tool. There's no summary string here (the processor node is the one that gives you those).
Install
Part of ComfyUI-NovelAI-GENYTOOLS. ComfyUI Manager, search "ComfyUI-NovelAI-GENYTOOLS", or:
cd ComfyUI/custom_nodes
git clone https://github.com/XTOGENY/ComfyUI-NovelAI-GENYTOOLS.git
pip install -r ComfyUI-NovelAI-GENYTOOLS/requirements.txt
Restart and set your NovelAI Persistent API Token under NovelAI 2.0: Persistent API Token in settings. No models to download, nothing to configure. And no Anlas - this node just rearranges text.
Gotchas
- You can't feed it a raw string; it only accepts the typed
NAI_POSITIVE_PROMPTport. If you have a plain text box you want to merge, run it throughNAIPositivePromptV2first. - Long chains of joins get visually noisy. Once a chain is stable, collapse it into a single prompt node with the combined text - the canvas reads better and nothing about the output changes.
- The separator is easy to forget when you convert it to an input. An empty separator produces
a+bjammed together, which is a silently wrong prompt. Keep it", ".
Inputs (3)
| Name | Type | Default | Description |
|---|---|---|---|
| a | NAI_POSITIVE_PROMPT | — | |
| b | NAI_POSITIVE_PROMPT | — | |
| separator | STRING | , | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| 正面提示词 | NAI_POSITIVE_PROMPT | — |