Setting Up ComfyUI for Local Image Generation

LocalRPG can generate scene illustrations, NPC portraits, and item icons directly as your tabletop campaign unfolds. For high-quality offline image generation, LocalRPG connects directly to ComfyUI over its local HTTP API on port 8188.

1. Hardware Requirements

Image synthesis is GPU-intensive and benefits heavily from dedicated VRAM:

Model Architecture	VRAM	Recommended GPUs
Stable Diffusion 1.5	4–6 GB	NVIDIA GTX 1060, RTX 2060, Apple Silicon (8 GB+)
SDXL (Recommended)	8–12 GB	NVIDIA RTX 3060, RTX 4060, Apple Silicon (16 GB+)
Flux.1 / SD3	12–24 GB	NVIDIA RTX 3090, RTX 4080/4090, Apple Silicon (32 GB+)

2. Running ComfyUI

Option A: Standalone Python Setup

Clone and install ComfyUI according to the official documentation, then launch it allowing local network listening:

python main.py --listen 127.0.0.1 --port 8188

Ensure your target checkpoint (such as sd_xl_base_1.0.safetensors or a fantasy fine-tune) is placed inside ComfyUI/models/checkpoints/.

Option B: Running via Docker

The community yanwk/comfyui-boot image installs ComfyUI at /root/ComfyUI, so checkpoints belong in /root/ComfyUI/models/checkpoints and generated images land in /root/ComfyUI/output.

docker run -d \
  --name comfyui \
  --restart always \
  --gpus all \
  -p 8188:8188 \
  -v ~/comfyui/models:/root/ComfyUI/models \
  -v ~/comfyui/output:/root/ComfyUI/output \
  yanwk/comfyui-boot:cu130-slim-v2

The image publishes no latest tag, so pick the variant that matches your hardware: cu130-slim-v2 for a current NVIDIA driver, cu126-slim for an older CUDA driver, rocm7 for AMD on Linux, and cpu when there is no GPU. The CUDA tags require the NVIDIA Container Toolkit on the host.

Docker Compose (docker-compose.yml)

services:
  comfyui:
    image: yanwk/comfyui-boot:cu130-slim-v2 # or cu126-slim, rocm7, cpu
    container_name: comfyui
    restart: always
    ports:
      - "127.0.0.1:8188:8188"
    volumes:
      - ~/comfyui/models:/root/ComfyUI/models
      - ~/comfyui/output:/root/ComfyUI/output
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]

Drop the deploy block when you use a CPU or ROCm image.

3. Verify Server Health

Test that ComfyUI is active and serving its API:

curl -s http://127.0.0.1:8188/system_stats

Verify that the response returns system metadata and GPU devices:

{"system": {"os": "linux", "python_version": "..."}, "devices": [{"name": "NVIDIA GeForce RTX 4090", "vram_total": 25757220864}]}

4. Connecting in LocalRPG

Via Settings Studio (GUI)

1. Open Settings Studio -> Media -> Image.
2. Click Load Image Preset… and select ComfyUI.
3. Confirm the Endpoint is set to http://127.0.0.1:8188.
4. Click Save Settings.
5. In your campaign or Codex entity view, click Generate Image to request an illustration.

Via Configuration File (config.yaml or localrpg.yaml)

media:
  image:
    type: http
    endpoint: "http://127.0.0.1:8188"

5. Alternatives: Automatic1111 & Stable Diffusion WebUI

If you use Automatic1111 or SD-WebUI-Forge, launch the web UI with the --api argument (such as COMMANDLINE_ARGS="--api"):

media:
  image:
    type: http
    endpoint: "http://127.0.0.1:7860/sdapi/v1/txt2img"

Select preset Automatic1111 in Settings Studio to connect.