Create ComfyUI Image workflow or similar

We're currently heavily leveraging the gemini image edit endpoint/openai image edit endpoint to extract clothing items from pictures that users upload. The pictures are sometimes clothes lying on the floor, sometimes someone wearing them, sometimes just clothes hanging in a wardrobe. We have a vision model that first identifies and extracts descriptions of each piece that it can see. Then we're using the openai image model where we give it the image and tell it to extract the jeans for example onto a clean background, product photography style. We're looking into the feasability of using our own servers to re-create the image extraction process. We need someone that has been toying and playing around with models like this for a long time, knows how to test/recreate and most importantly, knows how to deploy this at scale.