V2.0
Version 2 of Fine Anime Screencap is here and it comes with these improvements!
A new training technique was used
Improved screencap style
Improved color generation
Removed the blur produced in the generation (not to be confused with the blur produced by the sd itself)
Working with the weight from 0.6 to 1, use it according to your needs.
Even more compatible with many models
Recommended to use More details for better results.
V1.5
More images were added to the dataset.
Improved training technique
Fixed: sometimes generated wrong hairstyle
The recommended sampler is Euler a or DDIM.
Optional tag: anime screencap
Weight from 0.8 to 1
V1.
This Lycoris was created with the purpose of obtaining an anime style, but more fine and detailed with a thinner lineart style than the standard anime style, combine it with Loras of characters or landscapes, it works better with Loras that have a neutral style, although with Loras that were trained with screenshots it will also work although you won't notice much difference, optionally you can use the Lora of more details between 0.4 to 0.6 to obtain better results, use Highfix to obtain much more detailed images.
Recommendations:
Sampler: euler a
Steps: 25 or more
Size: 512x768 or 768x512
Weight: 1
Recommended use Anylora Checkpoint
example image generated with anylora without the style applied
now with the style applied
Share your generations and enjoy!