I'm testing out a lora to give a window view (from indoors) of a view of buildings, with a London character (both urban and residential). I think it probably needs more training and a wider data set. Please share feedback, thanks