Prompt Guide
The bundled prompting manual, as a node — write prompts the model's own docs say to
- guide
Prompt Guide is a glorified file reader, and glorified file readers are underrated. It loads the pack's bundled model-specific prompting guides - the docs the model vendors themselves ship - and outputs them as a single string. Pick a model and a mode, and you get the text-to-image or image-to-image guidance for that model, ready to read or ready to wire into an LLM.
Why this matters: every modern base model has opinions about how you should prompt it. Flux2's Qwen encoder wants natural structured instructions, the older SD family has its own conventions, and nobody remembers all of it. Instead of keeping a bookmark folder of markdown docs, you get them as a node output you can read in a Show Text node - or, more usefully, feed straight into the pack's Story Director and Asset Sheet Director via their guide_t2i / guide_i2i inputs, so the LLM writes prompts grounded in the model's own documentation rather than generic prose.
Inputs
model- auto-listed from the bundleddocs/models/folder:flux2,pony,sd15,sdxl,z-imageat the time of writing.mode-t2i,i2i, orboth(which returns the text-to-image and image-to-image guides together, labeled).strip_jsx(default on) - removes the JSX/React snippets and HTML tags embedded in some of the docs, leaving clean readable guidance. Leave it on unless you actually want the code.
The single output, guide, is one STRING. That's the whole node.
The honest caveat
This is only as useful as the docs bundled with your installed version. The model list is whatever folders exist in the pack's docs/models/ directory - if a new model comes out, you don't get a guide for it until the pack updates. And the guides describe the vendors' recommended prompting, which is a great baseline but not gospel; the community often converges on its own tweaks. Use it as grounding, not as a cage.
Install
Part of ComfyUI-ZFRNodes. Via ComfyUI Manager (search "ComfyUI-ZFRNodes") or:
cd /path/to/ComfyUI/custom_nodes
git clone https://github.com/zfrsgtcu/ComfyUI-ZFRNodes.git
Restart ComfyUI fully. No models to download, no API keys - the docs ship inside the repo. If you're hand-writing story prompts, wire guide into a Show Text node and skim it once; if you're letting an LLM write them, connect it to Story Director or Asset Sheet Director and let the docs do the prompting for you.
Inputs (3)
| Name | Type | Default | Description |
|---|---|---|---|
| model | COMBO | 5 options: flux2, pony, sd15, sdxl, z-image | |
| mode | COMBO | t2i | 3 options: t2i, i2i, both |
| strip_jsx | BOOLEAN | true | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| guide | STRING | — |