This is an experimental LoRA series trained on the Anima-preview model.

Version overview:

• data01-v1.0: Pure anime dataset, SDXL personally made mixed style dataset

• data02-v1.0: Another anime dataset, SDXL personally made mixed style dataset

• data03-v1.0: Based on data02 with additional partial 2.5D elements, can generate semi-realistic style with more sense of volume / depth appropriately

No specific trigger word required — load the LoRA and use descriptive prompts or Danbooru-style tags directly.

Common recommended settings (aligned with Anima official model card):

• Base Model → Anima

• LoRA Weight → 0.8–1.0 (0.8–0.9 usually most balanced)

• Sampler → er_sde / euler_a / Res Multistep

• Scheduler → Simple or SGM_Uniform

• Steps: 25–30

• CFG Scale: 4–5

Prompt tips:

- Start with quality tags like masterpiece, best quality, highres

- Negative prompt: In most cases, just add score_1, score_2, score_3 — this is often sufficient to avoid low-quality outputs. Adding more negative tags frequently impacts generation negatively (e.g., reduces variety or over-corrects details).

For more official Anima prompting details, check: https://huggingface.co/circlestone-labs/Anima