v2With version 2 I’ve expanded the dataset to closer to 800 images, monkeyed with some settings, bumped up the training resolution and added the trigger word: illustra. This is the final version for this Lora, the first of the illustra series I’m working on. Just my take on a few styles I like. Hope you guys enjoy.Prompts include with the images this time, though a most are just pulled from my earlier merges and don’t reflect training captions… or anima for that matter. Still, I’m pleased with the results.v1Brings a smooth comicish more western focused style to anima. Trained on a little shy of 400 sfw (not that Anima cares) images produced by zimage turbo.Example images were generated using wildcards hence no prompts. heres a few examples, but honestly, just follow animas normal prompting structure:masterpiece, best quality, score_7, safe, 1man, solo, elderly, wizard, long beard, staff, mountain peak, starry skymasterpiece, best quality, score_7, safe, 1woman, solo, pilot, flight suit, helmet under arm, hangar, jet backgroundmasterpiece, best quality, score_7, safe, 1man, solo, detective, fedora, trench coat, foggy street, lamplightmasterpiece, best quality, score_7, safe, 1woman, solo, flamboyant, extravagant gown, feathers, stage, spotlightNegative prompt:worst quality, low quality, score_1, score_2, score_3, blurry, jpeg artifacts, sepia, anime, amputee, missing limb, bad proportions, white background, boarders, boarder, comic panel,

Western Illustra

Larger and rejiggered dataset, higher training resolution, added trigger word: illustra.

Files with SHA256: 78c7dfc1ea29ea3aa6959d3957ba2dffced0fd5a32a9c7c674c3cd581f377f80