Demon Slayer / Kimetsu No Yaiba anime style LoRA
Making models can be expensive. Do you like what I do? Consider supporting me on Patreon 🅿️ or feel free to buy me a coffee ☕
Long time request, this is the style of the anime Demon Slayer, trained from screenshots of the first two seasons. Triggers with kimetsu no yaiba style at weight 1. Adding anime coloring helps too.
For the examples I used AnyLoRA which will give the most accurate result, but it might work on other anime models too, as long as they're not too much opinionated.
Be aware this is a style LoRA, so it requires character LoRA's or embeddings to be accurate (but it can do a fairly decent Tanjiro, as he was a big part of the dataset).
Enjoy.
How to use LoRA's in auto1111:
Update webui (use
git pulllike here or redownload it)Copy the file to
stable-diffusion-webui/models/loraSelect your LoRA like in this video
Make sure to change the weight according to the instructions (by default it's
:1)
Description
FAQ
Comments (20)
W style, ty
Dang, I should get you to do my LORA training now you ... SIGHS you're so much cooler at it than i am XD WHINES (and yes i'm downlaoding this to play with)
Don't give up.
@Lykon Thank you for that, i love how the AI community is way more supportive than any other one iv'e been through - and as someone who's got a design degree that means that AI is quite supportive :p
This is a nice style! I'll definitely try it. You say you still need the character loras… I've not seen loras for anyone other than Nezuko. Did I miss them somewhere? I really am looking for the Kamaboko Trio (Tanjirou, Inosuke, Zenitsu) most, but others too.
So this one seems to be able to do Tanjiro on its own. For other characters you'll have to make them. Shinobu's lora or TI should exist I think. I've seen it. Plus some characters might be famous enough in the base model.
@Lykon Okay, thanks for the info! I'll give it a try and see how it turns out.
Where's zenitsu? 🥺
He's in the training data a lot of times. Anyway this is not a characte lora. I still got his hairstyle from time to time.
Make Lora of Sono bisque doll style please
What does "offset" stand for?
it's just a versioning system I use.
@Lykon Alright, I was wondering if that was related to the noise offset setting in any way, though unfortunately I can't check training parameters in the Additional Networks extension
@OneMinuteDeen oh no, epi noise offset came out after. I was using this word before.
For some reason the eyes tend to look not so good, (i am using it in combination with a Mitsuri lora) otherwise its really good
are you doing full body pics? Use highres fix and adetailer if you do.
Upscale ur pics or inpaint the face with a higher denoise
img2img results are not likely? any advise?
very nice lora, I'm curious how you train the data, It seems that the output of lora is different from the style of screenshots
Do you plan on converting it to Pony?
Details
Files
kimetsu_no_yaiba_style_offset.safetensors
Mirrors
kimetsu_no_yaiba_style_offset.safetensors
demonSlayerKimetsuNo_offset.safetensors
kimetsu_no_yaiba_style_offset.safetensors
kimetsu_no_yaiba_style_offset.safetensors
kimetsu.safetensors
pytorch_lora_weights.safetensors
kimetsu_no_yaiba_style_offset.safetensors
kimetsu_no_yaiba_style.safetensors
dmnslyrkmtsnybnmstyllr.safetensors
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.