This one is pretty rough. Don Bluth animation is wild but also equally wildly annoying to train. The data sources were fairly broad so the style can drift a lot

Fingers can and will get fucky. Garbage in, garbage out.

Having more luck with mixed tags even though it's trained with captions. Blame it on the method I guess, I dont know. YMMV

Str 0.9 or 1

Cfg: 5 or 6

Anecdotal: Modifying shift higher might help as well (5 or so instead of 3)

I'm using quality tags to keep it from absolutely catching fire. The intention was to avoid disney but I really did a fucking shitty job on the captioning so it's possible half of them are disney anyways. That comes with its own set of biases (mouse ears lol) but you'll survive.

d0n-style, high quality studio animation, disney, masterpiece, absurdres, score_7

High recommend:

film grain or soft focus

mixed media detailed background

glitter, sparkles go fucking crazy and are super fun to use.

Negatives:

(photo, realistic, anime) worst quality, low quality, score_1, score_2, score_3, blurry, jpeg artifacts, sepia, low quality, worst quality, blurry, bad anatomy, extra limbs, deformed, watermark, text, signature, bareness, artifacts, hands, copyrights name, jpeg_artifacts, scan_artifacts, bad hands, missing fingers, extra digit, fewer digits, artistic error, ye-pop, deviantart, logo, patreon logo, child, loli, aged_down