One Piece Wano Style LoRA - V2 released
Making models can be expensive. Do you like what I do? Consider supporting me on Patreon 🅿️ or feel free to buy me a coffee ☕
After making a TI for the One Piece anime stile of the Wano saga, I decided to try with a model finetune using LoRA. I highly suggest you use Midnight Mixer Melt as base model. The model works fine with my Robin, Nami and Yamato embeddings. It can also make Luffy and Zoro without any embedding. I also highly suggest you use the WD VAE for more vibrant colors. Also remember to set CLIP skip to 2.
Weight:
between 1 and 0.6 for v2_offset
between 0.5 and 0.3 for v2_original
around 0.8 for v1
Trigger keywords are wanostyle (mandatory) and monkey d luffy for Luffy or roronoa zoro for Zoro. May also get nami and nico robin without embeddings.
Keep in mind this is LoRA, so you need to use it as such. It's not a standalone model and it's not a TI.
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 if needed (by default it's
:1)
In theory this can also be mixed with my Wano style embedding, but I never tried and shouldn't be necessary. Overall this one produces far cleaner results, with less strange stuff happening in the background.
Description
FAQ
Comments (34)
can you uppload ckpt?
this is not a model but a LoRA. They usually come in safetensor format or pt format.
kewl
"I also highly suggest you use the WD VAE for more vibrant colors. Also remember to set CLIP skip to 2."
1-Can someone explain to me if that WD VAE is a "BASE MODEL" OR THAT i NEED TO USE IT WITH "Midnight Mixer Melt"
2-Can someone help me understand what is CLIP SKIP TO 2?
The vae is not the base model.
Clip skip is a setting, as well as the vae. Midnight Mixer Melt is a base model that I duggest you use with this lora to achieve the best result.
@Lykon And where do I put the WD VAE? into the VAE folder? Even tho it´s a CKPT file?
@Lykon I have a question for you, is it possible to upload my photo somehow and make it in One Piece style? Do you know anything I could try?
@Mangazilla you coukd use controlnet or make a lora/embedding with it. Check my Jack Sparrow LoRA, I explained how there and even made Jack in One Piece style
@Lykon Amazing, how long did it take to train Jack Sparrow images into the base for turning him into anime style?
@Mangazilla I explained it on that page
@Mangazilla If you use AUTOMATIC 1111 webui, there is a folder called VAE on where you paste the WD vae. So, you go on the webui settings>>stable diffusion, select the appropriate VAE on the "SD VAE" field. About the clip skip, there is a field called "Clip skip" there too. In the version I use, that field's name includes a link to this wiki (github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features#clip-skip).
I know this comment was made months ago, but this is my answer anyway.
Would you be against training a checkpoint on 1995 anime series "Neon Genesis Evangelion"? It's a very popular series and a similar style checkpoint has already been done for Studio Ghibli. Would be amazing to see it for 199s Evangelion. Thanks!
Hello man it´s me again, I wanted to thank you for creating such an amazing Lora.
Also, is it possible to create shanks?
Not sure if it is with this lora or if you need to get a second one dedicated to Shanks or a similar character. Shanks wasn't in the dataset, so I'm not sure.
@Lykon Do you know some Lora dedicated to shanks?
@Mangazilla unfortunately not, sorry
有没有白胡子
Nice work! I use this Lora with many anime base models and it works consistently well. Can I ask how you collect and process the training data? I want to make a Wano-style Sanji (and other characters) but it does not give expected results probably due to a lack of training data. Thanks!
Hello, author. I have a question about your work. Each character does not contaminate each other directly。 "Can you share it with me? I am very interested. If you could reply to me, I would be really grateful."
uhm I'm not sure. What's the question? :D
@Lykon "No, it's your model that's too perfect. I'm amazed at how it manages to achieve such high accuracy!" So I'm consulting with you, how can I achieve the same results as you!
@873297465973 uhm generation data is visible on civitai. You're not getting good results? Try showing me an example of image you're generating and also your generation data (pnginfo) so I can try and help you
@Lykon I think he's trying to ask if you did anything special to train this model :)
@Blarg00n not sure, special as...?
Manga version pls
how do you get these type of colors ?
Hello, Lykon! Firstly, thank you for the model.
What mugiwaras are on the dataset?
I'm having trouble, The Vagina is missing. Any tips please?
wat
What is the difference between offset and original?
offset can be used at 1 weight. It's in the description.
还原度很高