Previously trained data set, the new F.1 ran a test and released it for everyone to play with. Here is a comparison of the output images under the same parameters