Node packVerificadoExterno

comfyui-character-suite

ComfyUI custom node suite for prompt engineering with persistent character and segment libraries.

Fuentegithub.com/DrkSun81/comfyui-character-suite

Responsable@DrkSun81D · adoptioncalidad

Perfil del paquete

Estado: Verificado
Publicador: @DrkSun81
Fuente: GitHub

Léeme

ComfyUI Character Suite

A node suite for prompt engineering with persistent character and segment libraries.


Version	`1.1.1`
Revision	`bb8b76d`
Released	2026-07-26

Changelog (recent)

1.1.1 (bb8b76d) — Fix CS_LoraStacker FLOAT "None" errors: strength/bypass widgets no longer use converted-widget hide (which skipped them in the prompt and shifted combo "None" into FLOAT fields). Visual collapse only; client-side float/bool coercion; server-side _as_float / _as_bool guards.
1.1.0 — LoRA Stacker node with CivitAI trigger-word picker and example workflow.
1.0.0 — Initial release (Character Expander, segments, Prompt Builder, web Character Manager).

Installation

cd ComfyUI/custom_nodes/
git clone https://github.com/DrkSun81/comfyui-character-suite
# or extract the zip directly into custom_nodes/

Restart ComfyUI. No pip dependencies required.

Windows note: Requires Python 3.9+ (tested on 3.11). The embedded Python included with the standalone ComfyUI build is fully supported.

Example Workflow

An example workflow is included at examples/character_suite_example.json.

Drag it into ComfyUI to load a fully wired setup covering all nodes:

{Avery} expansion through CharacterExpander
Segment injection via SegmentLoader → Prompt Builder (CharacterSuite)
CLIP encoding → KSampler → SaveImage
PromptCategorizer wired off the final positive prompt
SegmentSaver ready to capture the built prompt for reuse

Before running: update the checkpoint in the CheckpointLoaderSimple node to one you have installed. The default (illustrious/absoluteTerritory_v3.safetensors) is set as a placeholder.

Nodes

All nodes appear under the CharacterSuite category in the Add Node menu.

🎭 Character Expander

Class name: CharacterExpander

Finds {Name} tokens in your prompt and replaces them with that character's stored tags. The character's negative prompt is automatically collected and merged into the negative output.

Inputs:

Name	Type	Description
`positive_text`	STRING	Your prompt, e.g. `{Avery}, wearing armor, in a dungeon`
`negative_text`	STRING	Your base negative prompt
`merge_mode`	COMBO	`append` (default) or `prepend` — where character tags land

Outputs: positive, negative, expansion_log

Example:

Input positive:  {Avery}, wearing armor, in a dungeon
Output positive: 1girl, silver hair, heterochromia, red left eye, blue right eye,
                 athletic build, confident expression, wearing armor, in a dungeon

Supports multiple characters in one prompt: {Avery} and {Kira}

🗂️ Prompt Categorizer

Class name: PromptCategorizer

Splits a flat comma-separated prompt into labeled buckets using keyword heuristics.

Inputs:

Name	Type	Description
`prompt`	STRING	Any prompt string to analyze
`use_llm`	BOOLEAN	Send unclassified tags to a local LLM (default off)
`llm_api_url`	STRING	LM Studio / Ollama endpoint (optional)

Outputs: categorized_text (formatted summary) + one STRING output per category:

Output	Examples
`quality`	masterpiece, best quality, highres
`style`	anime, cel shading, watercolor
`character`	1girl, heterochromia, confident expression
`hair`	long silver hair, twintails
`clothing`	school uniform, thighhighs
`action`	running, holding sword
`pose`	cowboy shot, from above
`scene`	forest, sunset, rain
`lighting`	rim light, dramatic lighting
`misc`	everything else