A LoRA for the style of the artist Muchu Muchu (formerly BHM, which apparently stands for Bearhug Maniac) (Pixiv, Fanbox). He doesn't actually draw that much bearhug art anymore and has nothing to do with the fetish that shares the same abbreviation (whatever that actually stands for).

Like all my artist style models, I will remove this if requested by the original artist.

Anima_v1:

Updated style triggers to match Anima's format. Map from old to new:

formerly "bhm_new" => "@muchu muchu"

formerly "bhm_new, glossy" => "@bhm"

formerly "bhm_old" => "@trashup"

Concept and character triggers should be unchanged.

SDXL:

Primary trigger word is "bhm", additional style triggers are "bhm_new", "bhm_old", and "glossy". See the sample images for examples of what these do. Style & concept adherence seems to be best on base Illustrious, but more derived checkpoints like WAI-Illustrious give (IMO) nicer looking results.

Original character triggers:

Maid: "maid_tsubaki, black hair, short hair, bangs, maid, maid headdress, maid apron, frills, puffy sleeves, striped thighhighs, black dress, long sleeves,"

Nurse: "nurse_emily, red hair, green eyes, large breasts, nurse cap, nurse, white dress, bare shoulders, white thighhighs,"

Blitz: "blitz, blue eyes, short hair, blue hair, black hair, two-tone hair, large breasts, bodysuit, gloves, goggles on head, leotard, earrings, jewelry, jacket, skin tight,"

Jade: "jade, hair over one eye, short hair, multicolored hair, black hair, green hair, two-tone hair, green eyes, large breasts, bodysuit,"

Concepts:

"bearhug" - tends to make the male a faceless, textureless creature, as is BHM's style. This is already a Danbooru tag, so lowering the LoRA's strength should reduce the effect on the male.

"mucus" - a subcategory of what Danbooru tags as "slime (substance)". It's not quite as effective as I'd like, so I'll probably eventually do a stand alone LoRA.

"coil" - wraps the subject in a tentacle or noodlely creature.

There's several more that should be in there, I'll add them here as I test them

v1.0 - Illustrious: First pass at training on Illustrious, seems to have come out pretty well, though it's quite variable across different Illustrious derived checkpoints, especially for the older style.

NoobAI vPred-1.0 version - WIP.