LoRA for tentacles coming out of a book.
Anm_a1: Same dataset as IL_a1, just on Anima now. Works decently well, probably better than IL. Triggers no longer need extra weights. Natural language can help.
Main: tentacles from book
Book position: floating book, holding book, book on ground
Optional: glowing book
IL_a1: Alpha version for Illustrious, plan is to use it to augment data for a proper IL_v1 and Anima v1. You need to figuratively beat the model with a wrench (by literally turning up the weight of the trigger) to make it work with named characters. This is due to a lack of named characters that Illustrious knows in the training data. More epochs didn't help with that, just made eyes worse. Anyways, triggers:
Main: (tentacles from book:1.5), open book
Book options: (holding book:1.25), (floating book:1.25), book on ground, (one book:1.25), glowing book,
Exact weights may need increasing, particularly if they're: used late in the prompt (most generators weight tags by order by default AFAIK, most important tags go first), used in very long prompts with weight normalization enabled, or if other tags are weighted heavily. There's also some variation by checkpoint.
Description
Anm_a1: Same dataset as IL_a1, just on Anima now. Works decently well, probably better than IL. Triggers no longer need extra weights. Natural language can help.
Main: tentacles from book
Book position: floating book, holding book, book on ground
Optional: glowing book



