ComfyUI-OpenAI-Fashion-Bridge

ComfyUI OpenAI Fashion Bridge

This package gives you a simple test bridge from ComfyUI to OpenAI image generation / editing.

What is included

• openai_fashion_bridge.py — custom node
• requirements.txt
• workflows/01_openai_text_to_image.json
• workflows/02_openai_fabric_to_garment.json
• workflows/03_openai_fabric_replace.json
• workflows/04_openai_ootd.json
• workflows/05_openai_image_edit.json
• workflows/06_openai_front_to_back.json

1. Copy the custom node into ComfyUI

Copy the whole folder into:

ComfyUI/custom_nodes/ComfyUI-OpenAI-Fashion-Bridge

2. Install Python packages

Open terminal in your ComfyUI Python environment and run:

pip install -r ComfyUI/custom_nodes/ComfyUI-OpenAI-Fashion-Bridge/requirements.txt

3. Set your OpenAI API key

Windows PowerShell

setx OPENAI_API_KEY "YOUR_OPENAI_API_KEY"

Then fully close and reopen ComfyUI.

Or paste key directly in the node

You can also paste the key into the api_key field of the node for quick testing.

4. Restart ComfyUI

After restart, search for this node:

OpenAI Fashion Bridge

Category:

OpenAI/Fashion

5. Import a workflow

In ComfyUI:

1. Drag the JSON file into the canvas, or
2. Use Load / Import workflow

6. Replace the demo images

For workflows with LoadImage, change the image filenames to your own test files.

7. First test settings

Recommended first test:

• model: gpt-image-2
• if it fails, fallback to: gpt-image-1
• quality: medium
• size: 1024x1536

8. Task mapping

• text_to_image → /v1/images/generations
• fabric_to_garment → /v1/images/edits
• fabric_replace → /v1/images/edits
• ootd → /v1/images/edits
• image_edit → /v1/images/edits
• front_to_back → /v1/images/edits

9. Important note

This is a simple bridge for testing. It does not do mask-based editing. For fabric_replace, OOTD, and front_to_back, success depends heavily on the prompt and source images.

10. If you get an error

Read the second output of the node (debug_json) or open the ComfyUI console.

Most common fixes:

• wrong API key
• no prepaid balance
• gpt-image-2 not yet enabled for your account
• switch model to gpt-image-1
• image too large or request timed out