This LoRa is intended to help the user create anime illustrations with higher quality detail and lighting effects. In addition, it aims to make it easier to create images with various effects such as blurred background, light particles, waterdrop, etc.
NOTE: Feel free to try out this LoRa as much as you like, and if you wish, you are totally free to write your opinions and recommendations in the comments. It would be greatly appreciated.
------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Suggested Specifications:
Image size: 768x1152. If you use an Upscaler and want to save credits, you can use the image size of 512x768 at a 2x resolution (this gives good quality and is relatively cheap in credits).
Sampling Steps: 20 to 30.
VAE: Automatic.
Sampling Method: Euler A or Euler; both in "normal" mode (you can use any sampler you prefer, just make sure to use the TA information table for the best results).
CFG Scale: 7 to 10.
Upscaler: R-ESRGAN 4x+ Anime6B.
AD Detailer: Optional, based on your preference.
Base Model: The LoRa has been tested with several checkpoints. The one that achieves the best results is "VisualMix (Illustrious)-V.1" and “Amanatsu (Illustrious)". It also works well with “NTR MIX | illustrious-XL | Noob-XL - Xii”; and “NAIXL_MMMMix - v4.5”. However, you can use any checkpoint you wish and experiment as you wish.
Well, lastly, I hope this model helps you create the images you want and that you enjoy it. If you have any questions, feel free to write in the comments.
Description
This LoRa is intended to help the user create anime illustrations with higher quality detail and lighting effects. In addition, it aims to make it easier to create images with various effects such as blurred background, light particles, waterdrop, etc.
NOTE: Feel free to try out this LoRa as much as you like, and if you wish, you are totally free to write your opinions and recommendations in the comments. It would be greatly appreciated.
---------------------------------------------------------------------------------------------------------------------------------------
Suggested Specifications:
Image size: 768x1152. If you use an Upscaler and want to save credits, you can use the image size of 512x768 at a 2x resolution (this gives good quality and is relatively cheap in credits).
Sampling Steps: 20 to 30.
VAE: Automatic.
Sampling Method: Euler A or Euler; both in "normal" mode (you can use any sampler you prefer, just make sure to use the TA information table for the best results).
CFG Scale: 7 to 10.
Upscaler: R-ESRGAN 4x+ Anime6B.
AD Detailer: Optional, based on your preference.
Base Model: The LoRa has been tested with several checkpoints. The one that achieves the best results is "VisualMix (Illustrious)-V.1" and “Amanatsu (Illustrious)". It also works well with “NTR MIX | illustrious-XL | Noob-XL - Xii”; and “NAIXL_MMMMix - v4.5”. However, you can use any checkpoint you wish and experiment as you wish.
Well, lastly, I hope this model helps you create the images you want and that you enjoy it. If you have any questions, feel free to write in the comments.