Kani Biimu Style LoRA for Anima
An artist-style LoRA trained for Anima Base 1.0, based on mature-period artworks associated with Kani Biimu.
The LoRA mainly captures the overall illustration treatment, including facial rendering, eye design, linework, coloring, shading, and character presentation, while keeping character identity, clothing, pose, composition, and background reasonably controllable through normal prompts.
Trigger: @kanibiimuaf2020
Recommended LoRA strength: 0.8–1.0
Lower values can be used for a lighter style influence.
Training information
Base model: Anima Base 1.0
LoRA: Rank 32 / Alpha 32
Training resolution: 1024
Optimizer: ProdigyPlusScheduleFree
Text encoder / LLM adapter: frozen
Training images: 319
Validation images: 32
Released checkpoint: 2200 steps
Usage
Use the trigger together with ordinary Anima/Danbooru-style tags or natural-language prompts. Character, outfit, pose, expression, composition, and background should generally be described normally rather than relying on the LoRA itself.
Description
About this version
This is the first public baseline release. It provides a reasonably stable reproduction of the target style and normal prompt usability, but it is not intended to be the final version. Some loss of style consistency can still occur in highly complicated scenes or unusually demanding prompts.
FAQ
Comments (1)
You set your LoRA as a model checkpoint. In order to correct this, click on your model page on the three vertical points in the top right corner and then on Edit Model. There in the second line you can change the Type to LoRA and don't forget to click on Save at the bottom of the page. If you correct this, just answer to this comment and I will delete it.