ComfyUI Integration

For advanced AI workflows, OpenGPEX integrates with ComfyUI — a powerful node-based AI workflow system. This enables capabilities far beyond simple generation, including inpainting, style transfer, ControlNet, and custom pipelines.

What is ComfyUI?

ComfyUI is an open-source, node-graph-based interface for Stable Diffusion and other AI models. By connecting to a ComfyUI server, OpenGPEX gains access to:

• AI Inpainting: Content-aware fill for removing unwanted objects
• Style Transfer: Apply artistic styles to your images
• ControlNet: Use edge detection, depth maps, or pose estimation to guide generation
• AI Generation: Generate new images or variations with fine-grained control
• Custom Workflows: Import any ComfyUI workflow (.json) for specialized tasks

Setup

1. Open the ComfyUI Bridge drawer in the right sidebar
2. Configure your ComfyUI environment:
   • Local: Connect to a ComfyUI instance running on your machine (e.g., http://localhost:8188)
   • Remote: Connect to a GPU server running ComfyUI
3. Import workflow files (.json) — these define the node graphs to execute
4. Configure parameters exposed by the workflow (prompts, seeds, models, etc.)
5. Execute and receive results directly in the editor

Environment Management

You can save and manage multiple ComfyUI environments:

Switch between environments instantly depending on your needs — use local for quick tests, remote for heavy workloads.

Workflow Import

There are two ways to import ComfyUI workflows:

Method 1: Paste API JSON

1. Export a workflow in API format from ComfyUI's web interface (Save → API Format)
2. Open the ComfyUI Bridge settings → Workflows panel
3. Click Import from JSON and paste the API JSON content
4. The parser auto-detects:
   • Workflow mode: img2img (has LoadImage node) or txt2img (generation from scratch)
   • Input node: Auto-detected LoadImage node
   • Output node: Auto-detected SaveImage/PreviewImage node
5. Select which parameters to expose (prompts, seeds, model names, etc.)
6. Save — the workflow appears in your workflow list

Method 2: Import from Server History

1. Click Import from Server in the Workflows panel
2. The bridge connects to all your configured ComfyUI environments
3. It fetches the server's execution history and lists available workflows (deduplicated)
4. Browse the list and select which workflow(s) to import
5. The bridge auto-parses the workflow structure and detects parameters

Note: For server import to work, the ComfyUI instance must have execution history available. Run a workflow in ComfyUI at least once before importing.

Exposed Parameters

After import, you can configure which workflow parameters are directly editable in OpenGPEX:

Input Source (img2img)

For workflows that take an input image:

• Active Layer: Sends the currently selected layer's content
• Merged Frame: Sends the flattened/merged canvas result

Real-Time Progress

The ComfyUI Bridge maintains a WebSocket connection for live updates:

• Progress bar: See generation progress percentage in real-time
• Node status: Track which node is currently executing
• Queue position: Know where your job is in the queue
• Execution history: View all past results with their parameters

How Results Integrate

When ComfyUI returns results:

• Single image: Added to the canvas as a new trunk creation
• Multiple images: First image becomes a trunk creation, additional results are attached as branches (see Creations & Frames)
• Batch processing: Results are organized automatically in the TabDock

Privacy & Security

• ComfyUI Bridge connects only to servers you explicitly configure
• No data is sent to OpenGPEX servers — all communication is direct between your browser and your ComfyUI instance
• You have full control over which models and workflows are available

Next Steps

• LLM Integration — Simpler API-based image generation
• AI Tools — Client-side AI tools (no server needed)
• Masks & Non-Destructive Editing — Refine AI-generated results

Last updated: 2026-07-30