Not sure what you mean by "show your real face", but in general, it tries to stay true to the original image, meaning it should preserve all the features as is. The goal of it is to fix the image artifacts, not to change the face. Sometimes though, it can mess up (especially for rare features like nose rings or scars, etc) and it occasionally has a tendency to smooth the skin. The reason for that is, these are rare so they don't appear in our training datasets as often. Hopefully, as we get more data, these algorithms should become better overtime.