Here's the Annoying Bitchy MILF we all love to hate. Who none the less we would tap on a heart beat. Seems there were a couple of LORAs of her pre-DBZ self, but none of her MILF version and I am a sucker for MILFs even if they are the annoying bitchy type.
V2:
While V1 worked fine, as always laziness motivates me to find ways to work less. Thus I re cleaned the dataset, hunted for some a extra costume images, retagged and rebaked it So I'll also post the LOCON version in a while. I repeat the V1 is also nice so try whichever you want, this one should be a bit more rigid but should spit her costumes consistently, V1 on the other hands needs to be properly tagged. She also looks a bit MILFer in V1.
Works fine with: <lora:chichi_dbzV2:0.7> chichi_dbz, <outfit>
I added triggers for the following outfits:
Purple_Dress_Orange_Neckerchief: Variation of her normal attire in purple and red.
Chichi_Bikini_Armor: Chichi cosplaying as her self in her Bikini armor.
Chichi_Blue_China_Dress: Her DB tournament arc dress look alike. I don't remember her using it during her MILF days but i found images and i liked it. Her MILF version of the attire is pants less of course.
Yellow_China_Dress_Purple_Neckerchief: Variation of her normal attire in yellow and purple. This probably need to add Neckerchief as extra or in negative as most of the images didn't have it and it is iffy. Sadly couldn't make this one consistent as it seems to fight against the top of her dress and turns it into a ribbon.
As Always I'll add the dataset so people can check the tag summary and experiment.
Note: Dafuq that steam came from? I am as confused as you are.
V1:
The results are strange, not bad but strange, at times it seems to work more as a style Lora than a character Lora. I would recommend to use at a weight of .5 or higher. At lower weight it seems to get confused while drawing her eyes and you might get a sporadic blind Chichi.
Using with conjunction with the original chi-chi_\(dragon_ball\) danbooru tag gives strange results just like that tag does in general(Otherwise this LORA wouldn't be needed).
Seems to be rather forgiving and got nice results with several combinations.
Got good results with: <lora:Chichi_DBZ:0.5>, Chichi_DBZ, black_eyes, black_hair, hair_bun, big_breasts, jewelry, earrings, ponytail, chinese_clothes
Some good tags to use according to the dataset repetitions are:
jewelry
earrings
hair_bun
bangs
bracelet
chinese_clothes
dress
chi-chi_\(dragon_ball\)
ponytail
hoop_earrings
china_dress
Description
FAQ
Comments (20)
Found a lora that i wanted to use but since im using invokeai, is there any way of using lora on invoke ai ? :c
Seems they are still in the process of implementing it. https://github.com/invoke-ai/InvokeAI/pull/2712
If you really want to use LORAs right now, I would suggest to move to automatic1111
Does prompting allow for a more accurate chest size? This appears to have been trained on only art where the character's proportions are much different from any canon reference.
Should be able to, though honestly the dataset was handpicked with mostly medium to big sizes bust. So it might fight you a bit.
I normally try to train with fanart and doujins as much as possible. It is a lot less likely for Toei to try to wring my neck if i don't use their stuff directly. If i ever get sued, asking why they aren't suing fan art artists first might not be a good defense but it is something.
@knxo Has anyone been sued for making an AI model anyway? I thought it was like...an AI. I don't know, maybe I'm ignorant.
@TheFoodMage Not yet, yet being the keyword.
any dbz males soon? like roshi?
Unlikely, my motivation runs on lust.
@knxo of course, then maybe any milf version of Bulma?
@greekars22 Maybe in the the future, but to be honest I am not that fond of her Z or Super's short hair guise. I do like her appearance in the alternate future/Trunks timeline, but I doubt there's enough material for that one.
@knxo I'm not sure you are understanding where he's going with this question
@fr34ky You mean RoshixChichi porn? or ChichixBulma porn? That just makes him a respectable man of culture. Still doesn't change my motivations though. I just don't like Bulma with short hair and while i appreciate giving Master Roshi some fun, some generic old dude can fill the spot until someone with more motivation makes a lora of him.
@knxo lol, ok I just meant that the thought of Roshi banging all the DBZ girls should be a 'lust motivator'. Good LORA thought, keep it up.
@fr34ky It is nice, but a good part of making a LORA is selecting and cleaning the dataset. I may like the idea of Roshi banging all the DB girls but not enough to try to see thousands of his images and do light photoshop on his dong.
what models did u use in examples?
The lora was Trained on AnythingV4.5 The images were made using the same.
@knxo hmm, weird, im using anything v4.5 and trying to replicate the images with anything v3.0 vae, and doesn't seem to work.
@joule32 If you are trying a 1:1 recreation, the model is the safetensor pruned, checksum 6e430eb514 https://huggingface.co/andite/anything-v4.0/blob/main/anything-v4.5-pruned.safetensors Other than that i use the NAI default setup configuration values.
Bulma need!!
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.