These loras were trained on about 51-150 screenshots from the 1995 Ghost in the Shell movie. This is a lora to capture the artstyle of GITS, but you can prompt the major using "kusanagi motoko". I included "gits1995" as a tag in the loras but I don't believe it does anything special.
These loras were trained on Anima 1.0, Noob 1.0 V-Pred, Zootlustrious V2, Illustrious v0.1 and Autism Pony. I haven't tested it with anything else.
To get the results you see in the images above, these are the settings I used.
Hands can be pretty crummy in the pony lora, fyi.
Description
This version was trained on Anima 1.0 using 52 screenshot images manually cleaned up and sharpen for clarity.
For this version I had to completely set up a new dataset because Anima was picking up the poor quality of the old dataset that SDXL ignored. This time around I paid very close attention to every detail to make sure the lora captured the style of the movie as best as it could. The dataset is about 80% the same as the ones from XL. I decided to add more aesthetic portraits and aesthetic backgrounds for the remaining 20%.
You can use the trigger word gits1995 to help with style consistency. You can prompt for kusanagi motoko, as well as her form from the end of the movie by using tag "kusanagi motoko peak". The other characters are hit or miss. I mainly trained for the style and kusanagi motoko.