PromptStyler is a custom ComfyUI node that helps you keep your prompts consistent by applying a selected “style template” to your positive prompt, then outputting CONDITIONING for KSampler.

If you are new to ComfyUI: you can think of CONDITIONING as the “encoded text prompt” that the sampler uses. PromptStyler builds a styled prompt string, encodes it with the model’s text encoder (CLIP), and passes the result to KSampler.

What This Node Does (Exact Behavior)

Node name in ComfyUI:

PromptStyler: Prompt -> Conditioning (Style Picker)

Step-by-step:

You type a normal prompt in prompt (your subject + what you want).
You pick one style from the dropdown (the node loads style templates from JSON packs).
If apply_style is enabled:
- It builds a new prompt: style.prefix + your prompt + style.suffix
- For template_variant=default, it merges comma-separated phrases and de-dupes duplicates.
- For template_variant=flux_2_klein, it appends a short prose style guide after your prompt (when available).
It tokenizes and encodes the final prompt using the text_encoder (CLIP).
It outputs:
- positive (CONDITIONING) for KSampler
- styled_prompt (STRING) so you can see the exact final prompt

Important: styles are written to describe style only (medium/process/lighting/palette/texture/composition) and to avoid injecting unrelated objects that you didn’t ask for.

Install (Beginner Friendly)

Copy or clone this repo into:

ComfyUI/custom_nodes/ComfyUI_PromptStyler/

Restart ComfyUI (or use Reload custom nodes).
Confirm the node exists by searching for PromptStyler in the node menu.

How To Use (Recommended Wiring)

Typical workflow:

Add CheckpointLoaderSimple
Add PromptStyler: Prompt -> Conditioning (Style Picker)
Add KSampler

Connect:

CheckpointLoaderSimple.CLIP -> PromptStyler.text_encoder
Your text -> PromptStyler.prompt
PromptStyler.positive -> KSampler.positive

Optional (debugging):

Use the styled_prompt output to verify that the style template is doing what you expect.

Inputs / Outputs (Explained)

Inputs:

prompt (STRING): your prompt text (subject/action/attributes). Works best with comma+space phrases.
apply_style (BOOLEAN): when off, PromptStyler passes your prompt through unchanged (still outputs conditioning).
style (dropdown): pick one style. Dropdown entries are Category | Name | id.
template_variant:
- default: uses comma+space “phrase tokens” (best for SD-like prompt token lists).
- flux_2_klein: uses prose guidance when present (helpful for Flux-like models).
style_id_override (STRING): advanced. If set, the node uses that style id directly (useful right after adding a new style).
text_encoder (CLIP): wire this from your checkpoint/model loader.

Outputs:

positive (CONDITIONING): connect to KSampler positive.
styled_prompt (STRING): the final prompt string used for conditioning.

Style Library (Where Styles Live)

Styles live in JSON packs:

styles/packs/*.json (merged at runtime in filename order)
Legacy fallback (only used if packs are missing): styles/styles_v1.json

Pack file naming:

Numeric prefixes control merge order (example: 10_cinema.json loads before 20_photography.json).

Style entry schema:

{
  "id": "unique_snake_case",
  "name": "Human Name",
  "category": "Category/Subcategory",
  "default": { "prefix": "...", "suffix": "..." },
  "models": { "flux_2_klein": { "prefix": "", "suffix": "..." } },
  "tags": ["tag1", "tag2"]
}

Adding Your Own Styles (No Coding Required)

This repo includes a helper tool that writes a custom pack file (gitignored by default):

Default user pack: styles/packs/99_user_custom.json

Wizard (recommended):

C:\Comfyui\python_embeded\python.exe .\tools\add_styles.py wizard

After adding a style:

Copy the printed id
Paste it into style_id_override to use it immediately (no ComfyUI reload required)

List categories:

C:\Comfyui\python_embeded\python.exe .\tools\add_styles.py categories

Add a style directly:

C:\Comfyui\python_embeded\python.exe .\tools\add_styles.py add --name "Ink Noir (Modern)" --category "Fine Art" --core "ink wash, chiaroscuro, high contrast" --details "paper texture, soft bleed, dramatic mood" --tags "ink,noir"

Prompting Tips (So Styles Stay Accurate)

Put subjects/objects in your prompt (the user input), not in the style template.
Write prompts as comma+space phrases: ", " is the splitter used by the node’s default template mode.
If a style is explicitly specific (its name is a scene/subject), it may include those nouns. Otherwise templates aim to stay subject-agnostic.

Validate / Audit (Recommended Before Editing Packs)

Validate required fields and uniqueness:

C:\Comfyui\python_embeded\python.exe .\tools\validate_styles.py

Optional consistency audit (also flags banned gear terms):

C:\Comfyui\python_embeded\python.exe .\tools\audit_styles.py

Regenerate Curated Packs (For Maintainers)

Curated packs are generated by:

C:\Comfyui\python_embeded\python.exe .\tools\generate_style_packs.py

After regenerating packs, restart ComfyUI (or reload custom nodes) to refresh the dropdown.

Troubleshooting

Node doesn’t show up:
- Verify the folder path is exactly ComfyUI/custom_nodes/ComfyUI_PromptStyler/
- Restart ComfyUI or use Reload custom nodes
Styles missing or dropdown says “(no styles found)”:
- Check that styles/packs/*.json exists and is valid JSON
- Run .\tools\validate_styles.py
A style feels like it “adds extra objects”:
- Try template_variant=default first (token style)
- Keep your subject strongly stated in the user prompt
- If you find a template adding unwanted nouns, it should be fixed in the pack (PRs welcome)

Versioning / GitHub Releases (Preparing v0.3.0)

Source of truth:

Version: VERSION (mirrored in __init__.py as __version__)
Changelog: CHANGELOG.md (Keep a Changelog)

Suggested release checklist:

Run .\tools\validate_styles.py and .\tools\audit_styles.py (include output in your PR).
Ensure README.md describes new behavior/inputs.
Ensure CHANGELOG.md has the release notes under the new version.
Push to GitHub and let Release Please create the release PR (see RELEASING.md).

README.md Unescape Escape

ComfyUI PromptStyler