---zimage base - turbo
20+ step, around 6 cfg

Tested on EULER, Simple. (dpmpp_2m_sde + sgm_uniform is good too)

main trigger: ruikirik woman, (just use woman would give better result & more flexible)

main style tag: kimono, rope belt, (normally use for negative if you want to change clothes)

style1 tag: green hair, hair bow, fox mask, gloves, (red gloves, fingerless gloves, )

style2 tag: beanie, headphones, gloves, fingerless gloves,

Works on turbo, same thing with 8 step, 1 cfg. (less look like, weaker than base) not testting too much.

---ill

default look will just be kiriko face and hairstyle.
you can add: Japanese clothes + helmet + red leggings to get the in-game look, there are no in-game shoes, it's not trained in the Lora.

Using WAI-NSFW-illustrious-SDXL for best result

Tested on:
DPM++ 2M SDE Karras

896 * 1152

CFG 6

Hires. fix by 2 time

you will need Adetailer for full-body, but if used in a close shot it might show smaller eyes.

Tagged with LLM, clothes should be fairly tagged out, much better than before. the default clothes are just some random clothes.

Seem even now, the eyes are still smaller and the face isn't as round as in the CG, but it seems like it is caused by its face always changing in the video :-? (some of the eyes are bigger, sometimes the face is longer)

I made a discord server, it's in my profile, if you want to chat, share some ideas, share some images, or if you want me to do some Lora, feel free to join.

v.11 since has been learning a lot a making Lora here is an updated version, V11 able to change to most of the clouth, tho it might still hold the style.

more step, adding the trigger word Tongue, and more glowing eyes image in there should help a lot if you want to get that glowing blue eyes as what was in the CG.

all image is still coming from the animated short, so her face should stay the same, but because you can move her more, there might still be a difference,

V.10

Over all better than V.1 but not sure is better than V.2 because v2 still get a better full body and better fit to all checkpoint model.

I think this going to be the final version of this Kiriko one, this version should be the perfect version if you want to make a CG like Kiriko.

The good part is detail is great, make a face you like, a pose you like, high res, those Kiriko Shinning eyes, etc. Working with a full body, You can change whatever the lower dress you want.

The bad part: I gave it a few tries, seems like her upper dougi is unchangeable like other clothes can't override it. No lower part in the training data, So part of the lower leggings and sneakers won't be what is like in the game or CG and I have no idea how to only get the shoes in there.

Give it a try, share some posts, and hope you guys can teach me something from your post. :D

This is my second Lora hope I am getting better at this.

This does not have any anime pictures in the training data, so I hope this will work better if you want to make realistic or 3D version pictures.

This is training basic on the Kiriko Animated Short and a few in-game screenshots for full-body ones, but mainly upper-body pictures.

Hope you can get a close one as to what Kiriko CG was showing, and that is the main reason why am training this Lora.

V.1 and V2 is different things, V1 more like CG, V2 better fit everything.

see vision detail for more inofrmaiton.

Working best between 0.4-0.7 depending on model

Leave a post if you made one you like :3

Any suggestions are welcome too :D

Search AI Models & Addons

very few results