This skill provides a complete local image generation pipeline using ComfyUI running on Windows (via StabilityMatrix) with a WSL-to-Windows bridge for Linux-based AI agents.
[OpenClaw/WSL] --bridge--> [ComfyUI/Windows Host] --GPU--> [Images]
|
[StabilityMatrix]
Recommended: Use StabilityMatrix
Manual Installation Alternative:
# On Windows, clone ComfyUI
git clone https://github.com/comfyanonymous/ComfyUI
cd ComfyUI
# Install dependencies per ComfyUI docs
The bridge connects WSL agents to Windows ComfyUI:
Bridge Script Location: scripts/comfy-bridge/comfy-bridge.sh
Key Configuration:
COMFY_HOST="192.168.4.95" # Your Windows IP (see below)
COMFY_PORT=8188
Finding Your Windows IP:
# In Windows PowerShell
Get-NetIPAddress -AddressFamily IPv4 -InterfaceAlias 'Wi-Fi*'
CRITICAL: WSL localhost does NOT map to Windows localhost. You must use the Windows IP address.
Recommended Models (in order):
| Model | Size | Notes |
|---|---|---|
| ------- | ------ | ------- |
| SDXL 1.0 | 6.5 GB | Default, reliable |
| JuggernautXL | 6.6 GB | Good alternative |
| SD 3.5 Medium | 10.8 GB | ⚠️ Experimental, needs 16GB+ VRAM |
Installation via ComfyUI Manager:
Manual Download:
Place .safetensors files in:
C:\StabilityMatrix\Data\Packages\ComfyUI\models\checkpoints\
# Check ComfyUI status
./scripts/comfy-bridge/comfy-bridge.sh check
# Launch ComfyUI (if needed)
./scripts/comfy-bridge/comfy-bridge.sh launch
# Generate image with SDXL (default)
./scripts/comfy-bridge/comfy-bridge.sh generate "A sunset over mountains"
# Generate image with JuggernautXL
./scripts/comfy-bridge/comfy-bridge.sh juggernaut "A sunset over mountains"
# List available models
./scripts/comfy-bridge/comfy-bridge.sh models
# List output images
./scripts/comfy-bridge/comfy-bridge.sh outputs
Daily cron that generates historical event images:
Setup:
# The workflow is at: scripts/on-this-day/on-this-day.sh
# Test event fetching
./scripts/on-this-day/on-this-day.sh test
# Run full workflow
./scripts/on-this-day/on-this-day.sh run
Cron Job:
Output Location:
C:\StabilityMatrix\Data\Images\Text2Img\
The workflow can post to Discord:
# Post image to Discord (use message tool with filePath)
./scripts/comfy-bridge/comfy-bridge.sh outputs
# Then use Discord API or message tool to send
--listen 0.0.0.0 --port 8188comfy-workflow/
├── SKILL.md # This file
├── scripts/
│ ├── comfy-bridge/
│ │ └── comfy-bridge.sh # Main bridge script
│ └── on-this-day/
│ └── on-this-day.sh # Daily image workflow
└── references/
└── SETUP.md # Detailed setup guide
| Use Case | Recommended Model |
|---|---|
| ---------- | ------------------ |
| Daily automation | SDXL (fast, reliable) |
| Photorealistic | JuggernautXL |
| Creative/artistic | SDXL + custom prompts |
| Historical scenes | SDXL |
| ⚠️ High detail (16GB+ VRAM) | SD 3.5 |
共 1 个版本