Deltarune XL + 1.5 LoRA版本Rework XL v1.0 (ID: 351620)

Deltarune LORAs for SDXL, Pony, and SD 1.5

SDXL:

• Trained on Animagine and Pony models

• Covers 14 Deltarune characters with variations: Kris, Noelle, Susie (light/dark world versions), Jevil, Ralsei (Ch.1/2), Berdly (light/dark), Spade King, Lancer, Cyber Queen, Rouxls, Spamton (+Neo), Swatchling, Tasque Manager, Rudolph

Prompting Tips:

• Place trigger words at beginning of prompt (except Pony model)

• Pony model trained with shuffling, so you can use trigger words in different places with higher success

• Additional details help (e.g., Kris's sweater type)

• Caution with "masterpiece, best quality" in Animagine models, can cause gender bend and NSFW

Legacy Models:

• Characters may struggle to interact unless similar

• Can influence output style minimally

Rework Information:

• See "about this version" section for specifics, same goes for other models

Pony Model:

• Slightly undertrained, but sufficient for most characters

SD1.5 Model:

• Lower quality than SDXL, requires more detailed prompting

• Limited comparison data with other 1.5 models

描述:

Training from scratch with more data that includes multiple characters. This version is undertrained a bit, so there could be inaccuracies with the characters, easy to fix with additional details. Legacy versions are better at this, but it is the most flexible so far.

Additionally, it is possible to generate more than one characters, but it's still not yet consistent and the characters can blend onto each other - even with something small as a 'hug'. And 'kfcgang' (because I've noticed amount of fan art on this) and 'deltarune' are new trigger words, although it isn't really all necessary for the outputs and can confuse the model instead.

训练词语: darkkris,lightkris,darksusie,lightsusie,darknoelle,lightnoelle,darkberdly,lightberdly,maskedralsei,ralsei,kingspade,queencyber,rouxls,spamton,spamtonneo,swatchling,tasquemanager,rudolphh,lancer,jevil