OpenGPEX: Browser Image Editor That Runs ComfyUI Workflows

OpenGPEX runs in the browser and connects directly to your own ComfyUI instance.

Generate, edit, export in one window

OpenGPEX is an open-source (GPL-3.0), browser-native image editor built around a ComfyUI integration. Instead of generating an image in ComfyUI, exporting it, opening a photo editor and re-importing, the editor connects to your running ComfyUI instance and turns any workflow into a generation button inside the layer stack:

1. Import a ComfyUI workflow JSON, or pull one from the server's generation history
2. Choose which parameters to expose: prompt, seed, model and other inputs become form fields
3. Hit Generate, and the result arrives as a new layer
4. Keep editing with regular tools: background removal, adjustments, masks and text
5. Export as PNG, WebP or TIFF with transparency

The editor is fully non-destructive: edits are stored as reversible metadata on top of the original image. It reads 10+ formats including 16-bit TIFF, RAW files (CR2/NEF/ARW/DNG), SVG and HEIC, and handles 100MP+ images in the browser. A nice touch for the ComfyUI crowd: drag in any image generated by ComfyUI and OpenGPEX reads the embedded workflow JSON, showing the full generation config and letting you download the workflow file.

Privacy-first and self-hosted

The ComfyUI integration talks to your own instance, so generation stays on your hardware. AI features beyond ComfyUI (background removal, SAM 2.1 segmentation, upscaling) all run client-side, and nothing uploads unless you opt into the optional cloud sync. The source is available at github.com/gpex-cloud/opengpex, everything is pluggable through its plugin system, and the author published a step-by-step walkthrough that builds a D&D character portrait from a ComfyUI workflow end to end.

For anyone who chains ComfyUI output into manual retouching, this removes the export-import loop entirely, and it works with any workflow you already have rather than a fixed set of features.