Style LORA for cartoon series Star vs The Forces of Evil, with knowledge of Star Butterfly and Marco Diaz's designs, and the overall style of the show as well.

Suggested Settings:

Remix first image in post for suggested models and settings to generate images close to the show's own style (and see Checkpoint section below for more informations).

Checkpoint Anima Aesthetic 1.0b is probably the best model to mix and match different artists and styles while still keeping design and proportions close to canon, even at lower strengths. It's sadly not available for online creation (at the time of writing) on CivitAI. Aesthetic v1.1 tends to let anime proportions influence the result a bit more (when not combined with a Turbo LORA, see below). Turbo v.1.1 is very aggressive in producing images that are close to the show's style.

If you're looking for on-model results my current suggestion is to combine Aesthetic v1.1 (or, better, Aesthetic 1.0b if available) with the LORA Expressive Accurate Turbo for Anima at 0.55-0.8. This combines the softer colors and lines of the Aesthetic model with Turbo's penchant for simple geometric shapes and flatter backgrounds.

Steps: 23-27

CFG: 3.5

Strength: 0.9-1.0 is usually fine when trying to generate images that look as much as possible like the show (see below), lower values allow to mix and match with different styles. Results depend a lot on the artists / LORAs used. 0.7-0.8 might be enough for some, others might require values as low as 0.3-0.4 (Star's face and hair tend to remain quite consistent even at very low strengths thanks to Anima's default knowledge of her design; Marco's might suffer more)

Artists: when used on its own, with no other LORAs or artists, this model alone can't properly fight the anime-style basis of Anima. As such using at least another style belonging to an artist with a "Western" art style is strongly suggested. I had good results using the built in artist tag (@inker comics:0.7-0.9~), or even better the artist's dedicated LORA at about 0.6-0.8

Anima documentation suggests that score tags can be dropped when using an Aesthetic checkpoint. Personally I think they help adhering more to the canon proportions of the characters, but testing has been very random so I can't be sure.