Version 3.0Reworked the workflow completely and switched to SD 1.5 with IC Light for totally different approach than before. Also changed the background removal tool for an even better mask. Please let me know about any issues you run into.Version 2.0Reworked a lot of stuff and the workflow looks a little more complex now, but the output quality improved drastically. I hope you like it and I am looking forward to your feedback and thoughts!How does this work? It takes your input image and generates a background with your input image with CosXL. But CosXL changes the input image, therefore its not as good. I am bypassing this by removing the generated subject from the generated image and editing the input image back on top with some slight adjustments.Known issue in this version is that the mask is not perfect all the time, not sure how to improve that bit at the moment unfortunately.Version 1.0The first release of my ComfyUI workflow for editing the background of existing product photos. Custom nodes can be found with ComfyUI Manager. Read the description of each node on how to use it if you run into issues! CosXL_edit Model is a must, others won't work!This Workflow will detect the subject and edit it into your prompt. It will also do basic color adjustments and is made of group nodes to reduce the spaghetti and keep it simple. This does not give you perfect results. Editing the final image output (before the compositing and color matching in ComfyUI happens) in Photoshop or similar software will give way better results. I just wanted to make this ComfyUI only.Prompting can be a little weird sometimes and you might get outputs that look like your inputs. Play around a little and I am sure it will work out some way or another!

Boto's Background Swapper

<ul><li>Improved generated output a lot with multiple new steps</li><li>Added inpainting for generated background image to get rid of any unwanted edges</li><li>Added a refining pass with SDXL to increase background image fidelity</li><li>Added multiple new options for more control over the process</li><li>Now using Ollama to detect the image subject and put it into the prompt automatically, can be bypassed and done manually</li></ul>

Files with SHA256: 663dbfcf041b03d534603c0e25f0796d56ba8a379a5046abc6a2f9bbaec7a8be