>>19749990
I shouldn't do this, but I have been trialling a workflow that combines controlnet and faceswap, but that'd a be a whole other thread because I would have to consistently swap between a normal workflow and an inpaining workflow, it shows good results but isn't testing a wide variety of contexts, has low rez outputs (upscale yourself, plenty of free services out there), AND I have to find pictures that reasonably match the body type of the subject.

Hence why I might do it after this thread.