V0.25 12-step is just... idk, unstable? it's clearly worse than v0.24 4-step at prompt adherence, this test here only kept increasing step counts by 4, while keeping everything else the same, you can see results yourself. This is feedback.

Dear creator please scrap whatever you used for creation of v0.25 and start from v0.24 again, i don't know anything about fine tuning but i know that something went downhill and that 0.25 is worse performing than v0.24 and is a downgrade instead update. <3 love you for your RDBT turbo checkpoints and loras as it is only viable way for me to even use Anima in 1st place, thank you for your work <3 i did this to give you feedback. Eng not 1st language so i can't really explain myself, i keep thumbs up v0.26 will came out fantastic.