After a lot of testing I came to a conclusion, that version 1.0 is overtrained. It changes the style of the image too much. So I decided to remake the LORA (version 2.0) with much larger dataset and lesser training. This time it does exactly what it is supposed to (enhance pespective shots) without changing the overall style too much.This LORA was trained mostly on images with perspective view (sexy poses), so it excels at elevating this aspect the most. But it also can work as a detailer/enhancer.The examples above are with LORA (first image) and without LORA (next image).I recommend using the trigger word "sexyper" to increase the effect even more.

Files with SHA256: 0791dd9e6d265f134b5e75a9fb7302f65a61fb2dad6216343f436c22d032e434