Style LoRA trained on a set of my highest quality generations of my OCs that were originally generated with various revisions of <a target="_blank" rel="ugc" href="https://civarchive.com/models/24350/perfectdeliberate?modelVersionId=2725374">https://civarchive.com/models/24350/perfectdeliberate?modelVersionId=2725374</a> and <a target="_blank" rel="ugc" href="https://civarchive.com/models/1839213/duchesse-style?modelVersionId=2173380">https://civarchive.com/models/1839213/duchesse-style?modelVersionId=2173380</a> It does unfortunately seem to have picked up a habit of gradient/two-tone hair from the training set: use strength of 0.6 or lower may suppress it, or specify the colors if you want to lean into it.So far I've found best results generating at around 2MP (I use 1920x1200 for the landscape images, 1200x1920 for portrait).

Files with SHA256: 2bb7bcb5a7593305fd0855188ad2487899b6c2ca12fce36a97c278cdeefbac9e