Discussions

When using a photograph, how do you keep the result close to the original while modifying certain parts?
The problem I run into is that when I use a landscape photograph, the AI generates a result that has nothing to do with it.