Experimental Anima Base v1 LoRA based on poderdohokuto (conto).
This LoRA specifically targets the simple, flat style from this artist.
Description
FAQ
Comments (2)
This lora is so good ngl. You can either use it as a style lora or lowering the strength to something around 0.3 and it becomes kind of a lineart stabilizer. And it's only 4mb too, how???
Training primarily on clean lineart is why this moves the model in that direction. This will introduce a bias towards monochrome, which can be somewhat mitigated by post-training cutting out some early blocks that are partially responsible for the coarse layout and lighting. This, paired with DIM8 (22MB), using FP8 data type halves the size down then approximately half again by cutting blocks.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.