Ear Fix Tsubaki DiT2
A specialized LoRA designed to improve anatomical consistency in characters with animal ears, specifically trained for the Tsubaki.2 (PixAI DiT.2) model.
This LoRA removes unwanted human ears when generating kemonomimi (animal-eared) characters, ensuring clean and natural results with only the intended animal ears visible.
Trained on a curated dataset of high-quality anime-style images featuring cat, fox, and wolf ears, with no visible human ears, this model helps eliminate one of the most common generation artifacts in anime diffusion models.
Features:
- Removes extra or human ears in animal-ear characters
- Improves anatomical accuracy
- Works seamlessly with Tsubaki.2 (DiT.2)
- Maintains original art style and character design
- Suitable for cat ears, fox ears, wolf ears, and similar
Usage Tips:
- Recommended weight: 0.6 – 0.8
- Use with prompts including animal ears (e.g., "cat ears", "fox ears")
- Optional negative prompt: "human ears, extra ears, double ears"
Trigger words:
ear_fix_tsubaki, earfix, tsubaki_fix
Designed for consistency, simplicity, and clean results.
