Finally been able to attempt my first patch of Lora training. This is just a raw sample comparison between the 2 models (with and without my test LORA). This is just to compare the different between the three. So I left all the bad images in there to see what's they're good and bad at.
So basically, "Illustrious" is the base model that I'm currently using, Anima is the model that I want to switch over to because it 'suppose' to have better knowledge of anime characters.
The problem is that I don't like how the base Anima look. My plan was to try to train my current style onto Anima, to make it copy that style while using still running Anima base.
Currently, the old Illustrious model still looks the best. It has that high end anime with film grains that I really like. Which make sense, since it's the result of a year of mixing and tweaking to get it look like it is today.
This is my first attempt to make a style LORA. To be honest it kinda sucks😅 It was done with 300 sample images. I will try to add in another 200 images and re-train when I have time, hopefully that will give better result. If that doesn't work, I need to maybe clean up my sample data and restrict the poses in order to keep the style more consistent.
Posted: July 13, 2026
🔑 MMDDYY ☝️