The Legend of Vox Machina 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 ☕
This is one of my favorite series. I decided to do this LoRA because I noticed that there is very little western animation style in any SD model.
Season 1 version should be used at around 0.6 weight using AnyLoRA with the trigger vox machina style. This one can currently do dragons, getting them right about 20% of the time.
I'll probably do a v2 once season 2 is over with stuff from there too.
UPDATE: made v2 based on the first 2 seasons. Use it at around weight 1 or even higher.
Keep in mind this is not a character LoRA, so it can't really reproduce characters from the show, but it can get pretty close with accurate prompting. This is very much intended for you to create your own characters. It should work on basically every model, but it's accurate only on AnyLoRA as far as I've seen, as that one tends to be more "style neutral".
First 3 and last 3 example of season 1 images are characters from my D&D group. I remade all of them for season 1+2 version.
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 (by default it's
:1which is usually too high)
Description
FAQ
Comments (17)
Wow, looking forward to trying it out! Thanks for sharing! (update, seems to be working for me now)
that's odd, I could get good results even on AOM2. Try to copy the generation data and replicate the settings.
@Lykon I tried again and was able to get decent results with a lot of different anime models, all using the Anything V4 VAE. Something must have just been weird on my end. Is that the VAE you would recommend for any models using this LoRA? Thank you very much.
@SnowSultan that depends on the model. If I need more saturated colors I go with the waifudiffusion one
Could you do the shibakumiko style? I don't know if you do those (pacman)
I actually really wanted this! Thanks a lot for providing it, hopefully you will keep showing love to this project.
of course I will
Please make Hyulla lora. He is the one who actually played around with AI to train it his own style xD so he wouldn't mind, plus his style is amazing!
Though not all of his works are on danbooru.. some older ones from Blade&Soul are on his naver blog, and some stuff on Twitter
And I do have few screenshots from his twitch livestreams that are not uploaded works, I could maybe send it somehow
Or train Lora myself, if I figure out how it works on Google colab
Would it be possible to make one based on the Campaign 2 Mighty Nein Intro? That style is so awesome. Reminds me of the cartoons when I was a kid.
Yeah I agree. Regarding the intro, there may not be enough data, so it may cone out overfitted. I'd have to look into that later after I relocate.
We need American Anime (DOTA Dragon's Blood, Vox Machina, Dragon Age Absolution, & etc) model/checkpoint.
made Dragon Age