Nivora (Wuthering Waves) - Anima

詳細

ファイルをダウンロード (1)

モデル説明

Nivora (Wuthering Waves) — Anima LoRA

A character LoRA for Nivora from Wuthering Waves, trained on the Anima base model.

✨ Trigger word

nivora

Put nivora near the front of your prompt. Her signature look (grey/silver hair, green eyes with teal under-eye marks, fur-trimmed hat with gold triangular ornaments, parallel hairclips, choker, fur-lined jacket) comes through from the trigger alone — you don't need to describe her manually.

But you can also add them

grey hair, green eyes with teal under-eye marks, fur-trimmed hat with gold triangular ornaments, parallel hairclips, choker, fur-lined jacket

🎨 Recommended usage

  • Base model: Anima (trained on anima-base-v1.0; works on Anima-family checkpoints such as anima_baseV10 / animayume / animaOfficial).

  • - LoRA strength: 0.8 – 1.0 (0.9 is a good default).

  • - Sampler / scheduler: dpmpp_2m_sde_gpu + beta57 (Euler / Rectified-Flow also fine).

  • - CFG: 4.0 – 4.5 · Steps: 30 – 40.

  • - Prompt tips: works best for portrait / upper-body. Example:

  • masterpiece, best quality, newest, highres, 1girl, nivora, solo,

  • looking at viewer, upper body, simple background, soft smile

  • - Add outfit/pose/background tags freely — her identity stays consistent because those were captioned during training (so they're changeable, not baked in).

🧠 Training details

  • Type: LoRA · Network dim/alpha: 16 / 16 · Precision: bf16

  • - Dataset: 34 curated solo images (from danbooru), auto-tagged with WD14, trigger-first captions, deduped & quality-gated

  • - Schedule: 10 epochs, ~1700 steps, resolution 768, AdamW8bit, unet_lr 5e-5

  • - Trainer: lora-scripts-next / SD-Trainer (Anima LoRA)

⚠️ Notes

  • Character © Kuro Games / Wuthering Waves. Trained on publicly-posted fan art for non-commercial, fan use.

  • - Not affiliated with the game or the original artists.

このモデルで生成された画像