Really poor results from generative fill
Hi you guys.
So, I have been trying out the new generative fill tool and I think it is really cool. Unfortunetly I feel like almost everything I add (except for probably 1 in every 200 items) comes out with a horrible result. I have watched a couple of videos on f.ex YouTube where I feel like the way they are able to select and replace items are doesn't really work for me. As an example: If I select over a piece of clothing and put in information to change the piece or for example a colour, the result that comes back out is consequently abselutely dreadful, twisted, and never heeds the command that it is given (as an example, when asked to change the color on a piece of clothing from black to red just using "red sweater" as an example, the three examples twists the image and still comes out as black).
It is kind of frustrating, because it seems like I am doing pretty much the same as in most of the videos, but it is almost as if it doesn't understand the language I write in (English for my part). Another strange thing is that if I ask the AI to change the color of say a t-shirt, simply marking the t-shirt area, what it will do is usually keep the colour or perhaps change the color, but completly disfigure the tshirt.
Does anyone have the same kind of experience and might have any thoughts about how I can change or improve this? Would love any sort of feedback.
