I have an image of an object I want to put in a setting, and I’ve explored a few different approaches and here are my initial impressions
- Generate the entire scene with object image and prompt as input (find that object changes too much)
- Outpainting (results are of inconsistent quality)
- Use preselected image as setting and add by Masking, cropping, relighting, adding dropshadow with opencv etc (when doing this semi-manually the results are quite good
So far 3) is by far showing the most promise, which makes me think if there is any tool or approach that does this specifically, vs me stacking together a bunch of different enhancement steps?