while Anima does have a decent understanding of artist styles, it almost always forces watermarks, artist signatures, etc. on the generated image, which are pratically impossible to get rid off, especially with low CFG.

in order to fix this issue, the LoRA was trained on a dataset cleaned from all text. in this process, some elements typical of the original style are sometimes reduced or lost, like excessive use of onomatopoeia, Ah's and Oh's, floating hearts, etc.

with that in mind, you can mix Anima's own understanding of the style to get a stronger effect, but this will introduce the watermark/signature issue. the LoRA itself does not require a trigger word.