sail-sg/EditAnything — reverse-engineered prompt

Reverse engineered prompt

GitHub

Build me a local image editing app where I can upload a photo, click on parts of it to segment them, and then change just those areas with a text prompt while keeping the rest of the picture in place. I want it to feel like an edit anything tool, so it should handle object level and part level edits, like changing clothes, hair, eyes, or swapping a person, and it should also support text based targeting like “dog head” or “bench” if possible.

Please include a simple browser UI with image upload, interactive click selection, mask preview, prompt box, and generated results. I also want a mode for cross image region drag or merge, so I can take a selected area from one image and blend it into another. If it fits this project, add sketch to image generation too, where I can draw a rough layout and guide the result with text.

Make it runnable locally, download or load the needed models, and use the repo demos as references for the main flows. Look up current docs online if you need to.

Want more depth? Deep Reverse