I constantly switch between base models and got tired of remembering which quality tags go with which models and what negatives are needed for what, so i merged a bunch of different ones over and over with assorted loras to get a consistent style with minimal prompting. this is the result of 5 checkpoints and 15 loras. best results ive found are with er_sde/simple 24steps with 4.2 cfg. but it is very flexible based off my testing. i dont use any quality or negative tags pretty much ever unless im targeting one specific thing.

Smoothbrain Prompting

i know the name says mk1 but this is like the 50th test, it has 5 checkpoints and 15 loras,

Files with SHA256: 8be2d7859a328b8ae135741e0034e2bf9a96f502cdff5a674c81d282c77bc62c