Lina Inverse Remake (Patreon)

Decided to completely redo the lora according to Lina Inverse, new optimal settings, a new model for learning and the already familiar Abyssorangemix3aom3 inference model, the results came out ... ambiguous, of course there are many more details, the lora has become more voluminous, but since the old "crutches" with resolution Art has not gone away, that is, there are also disadvantages, Lina became a little less like herself in the 90s, but this is all solved by increasing the weight of the model from the usual 0.6 to 0.8-0.9, and even 1 works well. I would like to completely redo it with new art, completely rewrite prompts, and I will get to that, but not today.

P.S. I learned something again, which makes me incredibly happy, learning from your mistakes is actually quite fun

P.P.S. The previous version is still quite working, the choice is yours