Sign In
homemodelsimagesvideos3D Modelshubsarticlescomicschallengesupdatesshop

WitchStyle – RPG Character Illustration Style

Updated: Sep 20, 2026

stylewitchfull bodycharacter designanimagame asset

Download

1 variant available

bf16 SafeTensor

WitchStyle_v1.safetensors

BF16, good balance • 87.61 MB

Verified:

Type
LoRA
Stats

79

Reviews

(14)

Published

Sep 20, 2026

Base Model

Anima

Training
Steps: 1,100
Epochs: 20
Usage Tips
Strength: 0.8
Hash
AutoV2
5796F83E7B
default creator card background decoration
Followers - 1

1

Likes - 14

14

License:

Anima

The 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

Text2Img_00005.png

WitchStyle – RPG Character Illustration LoRA for Anima

WitchStyle is a style LoRA trained for CircleStone Labs Anima Base v1.0. It is designed to generate full-body fantasy RPG character illustrations, especially witches, monster girls, female characters, elaborate costumes, and game character assets.

Trigger word

weiwitchstyle

  • Base model: Anima Base v1.0

  • LoRA strength: 0.6–1.0

  • Recommended starting strength: 0.8

  • Resolution: 768×1152, 832×1216, or similar portrait resolutions

  • Steps: 20–35 for Anima Base

  • CFG: 4–5

  • Sampler: er_sde, euler, or dpmpp_2m_sde_gpu

For Anima Turbo workflows, lower step and CFG settings may be used according to the base model recommendations.

Prompt example

masterpiece, best quality, score_7, safe, weiwitchstyle, 1girl, solo, full body, standing, fantasy witch, ornate black and red dress, long hair, detailed costume, white background, game character illustration

  • 0.5–0.65: Subtle style influence

  • 0.7–0.85: Recommended balance

  • 0.9–1.0: Stronger WitchStyle appearance

This LoRA primarily changes the illustration style. It is not intended to reproduce one specific character. Results may vary depending on the prompt, resolution, seed, and LoRA strength.

Training information

  • Training images: 55

  • Training steps: 1,100

  • Network rank: 32

  • Network alpha: 16

  • Training resolution: 768 with aspect-ratio bucketing

  • Trigger word: weiwitchstyle

Compatibility

This LoRA was trained specifically for the Anima architecture. It is not intended for SD 1.5, SDXL, Pony, Flux, or Z-Image.

Please follow the license and usage requirements of the Anima base model.