Sign In

Anime Lineart Style

102
830
20
Verified:
SafeTensor
Type
LoRA
Stats
830
Reviews
Published
Jul 6, 2023
Base Model
SD 1.5
Training
Steps: 39,000
Epochs: 1
Usage Tips
Clip Skip: 2
Hash
AutoV2
DDADDB711D

This is a style lora I created from 260 images generated using the 3moon anime_line model. I wanted to get the specific face style of the model, where the faces were longer with pointed chins. I also wanted to recreate the almost ethereal illustration design that the model was capable of.

I managed to get something similar enough to create intricate illustrations. However, the strength is a little too overpowering at full strength, so I recommend using 0.4-0.8 strength unless you are specifically looking for the highly saturated, vibrant colors of full strength. I included sample images of 1, 0.8, 0.6, and 0.4 strength so you could see the difference.

This is the design I'd hoped to achieve with my Foolkat_Anime_Merged_Model, but wasn't able to achieve. I decided to release it as a Lora so that it could be used for any anime model.

It does have issues generating hands, but that can be fixed through inpainting/img2img.

I included images of males in the training data, but I must not have included enough, because all of my attempts to create male images have resulted in female subjects. Hopefully, male characters can be created using character loras or using a different model than Foolkat_Anime_Merged_Model.

Sample images were created using FoolKat_Anime_Merged_Model version 2 for flat anime images.

All of the images I upload for this model are unedited and don't use any special extensions like ControlNet. I do not do any inpainting or img2img to correct errors. Why? Because you deserve to see how the model works without any touchup or special extensions like ControlNet. You can surely get better images by putting in the extra work, which I do for my personal images. The samples are just meant to showcase what the model can do raw with nothing but prompt engineering.