Extensions/ComfyUI-RPG-Characters
ComfyUI Extension

ComfyUI-RPG-Characters

Stylized RPG character prompt generator for ComfyUI. Supports standard and Ollama-based prompts, works with SD, SDXL, Flux, and more.

By lord-lethris·Created about a year ago·Updated 7 days ago· 7
lord-lethris/ComfyUI-RPG-Characters
Nodes6
On cloudLocal install
CategoryRPG, Prompting
Stars7
Updated7 days ago

Nodes (6)

Model Selector (Use Likeness)

An A/B switch for checkpoints — flip between your likeness model and your general one

RPG
RPG Prompt Concatenator

Merges your art and character prompts, then encodes them

RPG
Prompt → Conditioning

Two CLIP Text Encodes, one node — the bridge from raw text to conditioning

Prompting
Prompt Selector

A/B test your prompts — flip between standard and Ollama on one toggle

RPG
RPG Art Style Selector

Six art styles on a single dropdown — the style half of your RPG portraits

RPG
RPG Character Selector

A D&D character sheet that writes your SDXL prompt for you

RPG
Readme

ComfyUI-RPG-Characters

Current release: V2.5

A custom node for ComfyUI that generates stylized prompts for RPG characters. This node outputs both standard prompt formats and enhanced Ollama-style descriptive prompts ideal for extreme close-up portraits.

✅ Compatible with Stable Diffusion, SDXL, Flux, and any model that uses structured prompts.

The generator supports both classic AD&D / Fantasy and Sci-Fi / Cyberpunk character creation, with expanded classes, clothing styles, scenes, and art styles. V2.5 adds richer deterministic variation inside existing character choices and introduces a dedicated Pixel Art style. Sci-Fi characters can be created without augmentations, allowing fully augment-free characters as well as cybernetic ones.

✨ Features

  • 🎨 Choose from a variety of art styles:

    • Anime Style
    • Dark Fantasy
    • Realistic
    • Fantasy Illustration
    • Pixel Art
    • Sci-Fi / Cyberpunk

  • 🧠 Generates:

    • Standard Positive & Negative Prompts (for SD, SDXL, Flux)
    • Ollama-style Descriptive Prompt (text-to-text guidance)
    • Ollama-style Negative Prompt Instruction

  • 📸 Designed for 1024x1024 extreme close-up portraits

  • 📄 Fully structured to support character features:

    • Race, Ethnicity, Gender, Age, Class
    • Hair & Beard Styles and Colours
    • Clothing Style, Emotion, and Scene
    • Fantasy and Sci-Fi / Cyberpunk classes and clothing
    • General / Neutral, Fantasy, and Sci-Fi / Cyberpunk scenes

  • 🧬 DNA Locker (V2.4) — Deterministic character DNA with per-locus re-rolls, Master Seed control, and a 2D DNA Sculptor for blending the nearest three variants. Preview changes live, apply or cancel sculpting, and reproduce character variations from the same seed.

  • 🧬 Richer Character Variants (V2.5) — Existing character choices now contain more deterministic internal variation, including richer beard and hair-style descriptions plus colour, trim, material, and detail variants inside existing clothing styles.

  • 🖼️ Expanded Art Styles (V2.5) — Refined Fantasy Illustration prompting for a more classic tabletop RPG illustration look and replaced the overlapping Digital Painting style with dedicated Pixel Art prompting.

🆕 V2.5 Data & Art Style Refinements

V2.5 is a focused refinement release. It keeps the existing character-selection system and DNA Locker mechanics intact while making the underlying character data more visually varied and improving the art-style prompts.

🧬 Character Data Refinement

Existing user-facing choices have been enriched with deterministic internal prompt variants rather than adding large numbers of new dropdown options.

  • 🧔 Beard styles — richer variation in length, texture, grooming, and shape while preserving the existing beard-style choices.
  • 💇 Hair styles — restrained variation for generic styles, adding differences in length, texture, volume, and arrangement without changing distinctive named styles.
  • 👘 Clothing styles — existing Fantasy, tavern/service, and Sci-Fi clothing choices now include appropriate internal colour, trim, material, and accent variations.
  • 🧬 Tiefling pigmentation — expanded red-family skin-pigmentation variants and removed the human-skin fallback.
  • 🎲 Deterministic by design — these internal {A|B|C} variants are resolved through the existing character DNA system, so the variation remains reproducible.

The goal is simple: the dropdown describes the character choice; the DNA quietly determines the visual flavour.

⚠️ DNA is deterministic. Genetics are not.
Diffusion models can still interpret traits creatively, so deterministic prompt data does not guarantee an identical visual result across every model or checkpoint.

🎨 Art Style Refinement

V2.5 also refines the style prompts used by the RPG Art Style Selector:

  • 🏰 Fantasy Illustration — now explicitly targets classic tabletop fantasy RPG artwork, hand-painted illustration, expressive linework, visible brushwork, stylised heroic forms, and pulp-fantasy aesthetics rather than photographic realism.
  • 🕹️ Pixel Art — replaces the previous Digital Painting style with a genuinely different visual language: classic fantasy RPG pixel art, crisp hard-edged pixels, deliberate pixel clusters, limited palettes, dithering, sprite-art forms, and block-based shading.
  • 🎨 Digital Painting removed — Fantasy Illustration already covered the painterly/illustrated territory, so the overlapping style was replaced with Pixel Art.

V2.5 deliberately does not introduce new DNA mechanics, segmentation, regional conditioning, or a custom sampler. Those remain potential future architectural work.

🆕 V2.4 Workflow Highlights — DNA Locker

The AD&D Character Portrait Generator V2.4 introduces the 🧬 DNA Locker, a deterministic character-variation system that lets you experiment with a character's visual traits without losing combinations you like.

🧬 DNA Locker

The DNA Locker gives each generated character a reproducible set of DNA loci — such as skin tone, eye colour, hair colour, and other supported visual traits.

  • 🎲 Re-roll individual traits — keep rolling a specific locus until you find a combination you like.
  • 🎲 Re-roll the Master Seed — generate a completely new DNA combination across all visible loci.
  • 🧬 DNA Sculptor — blend the nearest three variants in 2D DNA space for finer control than simple random selection.
  • 👁️ Live feedback — DNA changes are reflected immediately in the Locker while experimenting.
  • ✓ Apply / ✕ Cancel — sculpt changes can be previewed live and either committed or discarded.
  • 🔢 Reproducible — the Master Seed allows the same DNA combination to be recreated.

The DNA Locker is designed to sit alongside the normal character-selection workflow: roll the dice, find a combination you like, sculpt it if you want more control, then run the workflow.

⚠️ DNA is deterministic. Genetics are not.
Diffusion models can occasionally interpret traits creatively, so the final image won't always perfectly match the DNA Locker. DNA is fragile. Handle with care. 🧬😏

Your character. Your DNA. Your dice.

🖼️ V2.4 Example Workflow

The repository includes a complete AD&D Character Portrait Generator V2.4 - Basic example showing the DNA Locker in use.

DNA Locker Highlight:
AD&D Character Portrait Generator V2.4 - DNA Locker

Workflow File:
AD&D Character Portrait Generator V2.4 - Basic.json

🆕 V2.3 Workflow Highlights

The AD&D Character Portrait Generator V2.3 workflow builds on the RPG character prompt system with several workflow improvements:

  • 🧩 Updated upscale stages to use the newer ComfyUI Subgraph system
  • 🎨 Added an Anime Prep stage to improve results when using likeness images with Anime-style generation
  • 🧙 Expanded AD&D / Fantasy character classes
  • 🤖 Added Sci-Fi / Cyberpunk character classes
  • 👕 Expanded Fantasy and Sci-Fi / Cyberpunk clothing styles
  • 🌍 Expanded scene selection with:
    • General / Neutral environments
    • AD&D / Fantasy environments and interiors
    • Sci-Fi / Cyberpunk environments and interiors
  • 🍺 Added fantasy tavern and inn environments, including Pub/Tavern Interior
  • 🧪 Added scientist, laboratory, corporate and research environments
  • 🧬 Sci-Fi characters are not required to have augmentations — augment-free characters are fully supported

The workflow remains designed for straightforward character creation while allowing the individual character, clothing, scene, and visual style to be mixed and matched.

📦 Install via ComfyUI Manager (Recommended 🎉)

The node is now officially listed in ComfyUI Manager!

To install:

  1. Launch ComfyUI and open Manager (via sidebar or custom_nodes menu).
  2. Go to the Install Custom Nodes tab.
  3. Search for: RPG-Characters
  4. Click Install
  5. Restart ComfyUI — you're ready to go!

🧱 Node Outputs

| Output Name | Description | |-------------------------------|-------------| | positive_prompt | Standard positive tag string | | negative_prompt | Standard negative tag string | | Ollama_Positive_Textbox_1 | Detailed descriptive prompt for LLMs or advanced generators | | Ollama_Negative_Textbox_1 | Instructional template to guide LLMs on what to exclude | | Ollama_Positive_Textbox_2 | Extra prompt input for LLMs or advanced generators that have a second Text Input|

📂 Example Files

All examples are in the Examples/ folder of this repository.

🧠 Standard Prompt Examples

Node Setup:
Standard Prompt Nodes
Workflow File:
RPG_Nodes_Normal.json

🧠 Ollama-Driven Prompt Examples

Node Setup:
Ollama Prompt Nodes
Workflow File:
RPG_Nodes_Ollama.json

⚔️ Comparison: Standard vs Ollama

Side-by-Side Visual:
Standard vs Ollama
Workflow File:
RPG_Nodes_Normal_Vs_Ollama.json

🔁 Prompt Switcher Example

Prompt Switch Node Example:
Switch Example
Workflow File:
RPG_Nodes_Normal_and_Ollama_With_Switch.json

🧩 Full Node Collection

A visual reference of the full node layout.
Full Node Collection

🛠️ Manual Installation (if needed)

Clone this repo into your ComfyUI custom_nodes folder:

git clone https://github.com/lord-lethris/ComfyUI-RPG-Characters.git

Restart ComfyUI after installation.

🧠 Usage

  1. Add the RPG Art Style Selector and RPG character nodes from the "RPG" category.
  2. Select the character traits, clothing style, scene, and art style you want.
  3. Mix Fantasy, Sci-Fi / Cyberpunk, and General / Neutral options as required.
  4. If using V2.4 or later, use the 🧬 DNA Locker to re-roll individual traits or the Master Seed, then optionally sculpt the DNA for finer control.
  5. Connect outputs as needed:
    • positive_prompt / negative_prompt → your SD/SDXL/Flux prompt node
    • Ollama_Positive_Textbox_1 / Ollama_Negative_Textbox_1 → LLM or prompt analysis/feedback tools

⚠️ Optional Ollama Support Requirements

To use the Ollama-driven prompt generation features in this node, you need to:

Without these installed and running, the Ollama prompt outputs in this node will not function.

Credit for the ComfyUI-Ollama node goes to Stav Sapir (stavsap).

🔮 Future Plans

  • Dynamic LLM integration for auto-generating character prompts
  • Additional art styles (e.g., steampunk, noir, and other distinct visual styles)
  • Further expansion of character classes, clothing styles, and environments
  • Additional workflow improvements and generation presets

🐾 Made by Lord Lethris

Featuring Belle the cat, who disapproves of bad prompts. 🐱