Hell's Paradise (地獄楽) 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 ☕
Should be used at around 1 weight using AnyLoRA with trigger hellsparadise style to improve the style.
Keep in mind this is a style LoRA, meaning it can get close to character but it's not aiming at doing them without a dedicated character LoRA/embedding.
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
:1)
Description
FAQ
Comments (13)
Jesus man , what the hell . the anime is not even done yet . don't get me wrong . I am impressed . is it possible for you to share a guide on how to train lora for motion lines or to indicate motion for character (if possible )
uhm I've never tried that myself, not sure I should write a guide. I suppose you just need good data for it.
@Lykon Seriously though, nobody does style LORA’s here better than you. Appreciate the hard work🙏
Nier Automata Anime please
ok
@Lykon thanks
Hi! Could you help me please? I am looking for information about style and concept loras. I am fairly confident in making character loras, but struggling with style and concept. Could you please suggest a tutorial where I can read about these, because I cant find anything myself. Thank you!
it's not much different. Just appropriate dataset and tagging. If you already do non-overfitted and flexible character loras (so NOT the type of lora that uses 1 tag only), then it's basically the same thing.
@Lykon What about the number of images? I usually work with around 100 images for a character lora. I would imagine concept loras would require more.
@Yuno I've done style loras with around 100 images (like my ff14 ones). For anime I usually use 800+
My One Piece Wano style lora being the only exception with 592 images (that dataset is a bit old), but it's still good.
@Lykon Last time I tried to do a concept lora I used around 300 images, but it turned out to be awful. I will try it again with more images, thank you! I usually do around 1500 training steps, for 800 images this might be too small, maybe around 8000? How many steps do you suggest?
@Yuno try 3000-4000
@Lykon Will try it thank you!
Details
Files
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.