Download
1 variant available
License:
AnimaThe Anima Model is licensed by CircleStone Labs LLC. Copyright CircleStone Labs LLC. IN NO EVENT SHALL CIRCLESTONE LABS LLC BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH USE OF THIS MODEL.
Built on NVIDIA Cosmos
## Overview
A style LoRA for Anima-based models, focused on chibi and super-deformed character illustrations.
It adds large-head/small-body proportions, bold and tidy outlines, bright flat colors, simple cel shading, and a clean sticker-like finish. It works especially well for full-body characters, idol costumes, expressive poses, and cheerful character art.
## Usage
Trigger word: habeli_sd_style
Recommended weight: 0.6-0.8
Example prompt:
habeli_sd_style, chibi, super deformed, full body, thick outline, flat color, simple cel shading, simple background, soft ground shadow
Start at 0.7 and adjust as needed. Lower weights preserve more of the base checkpoint's style, while higher weights emphasize the chibi proportions, bold outlines, and simplified coloring.
## Recommended settings
- Base model: Anima
- Recommended resolution: around 1024px
- Steps: 24-36
- CFG Scale: 4.0-6.0
- Shift: 8-12
- Sampler: DPM++ 2M
- Scheduler: Normal
## Compatibility
Trained on waiANIMA_v10Base10.
The uploaded sample was generated with anima_aestheticV11 at LoRA weight 0.7.
Compatibility with non-Anima architectures is not guaranteed.
## Training information
- Training images: 31
- Steps: 1,200
- Epochs: 5
- Network rank: 32
- Network alpha: 16
- Training resolution: 1024x1024 base with aspect-ratio bucketing
- Optimizer: AdamW8bit
- LR scheduler: Cosine
- Learning rate: 0.0001
- Precision: bf16
Results may vary depending on the checkpoint, prompt, seed, and LoRA weight.