Base model: anima-base-v1.0 (NOT SDXL / NOT Pony). Load with the Anima base model only.
A soft, low-saturation 2D illustration style trained on 23 hand-picked manga panels. Clean lineart, gentle flat shading, pale skin with warm blush. Best for quiet, emotional, slice-of-life scenes.
Recommended settings
- LoRA weight: 0.8
- Sampler: er_sde / simple, 25-30 steps, CFG 4.5
- Resolution: 832x1216
- Do NOT use hires-fix (it redraws the face and breaks the eye shape)
Prompt tips
- Always put the trigger @moeink first
- This LoRA carries a face — pair it with a character LoRA if you need a specific person
- Eyes: use "deep crimson eyes, gradient iris, vertical pupils" instead of plain "red eyes"
- Avoid "defined nose bridge" — it produces an oversized shiny nose
Negative prompts that matter
big nose, thick nose, nose highlight, rough lines, sketchy, heavy grain, overexposed, speech bubble, chinese text, japanese text, text, white box, watermark, loli, child, glowing eyes, bug eyes, worst quality, bad anatomy, extra fingers
Tested
- Scenes: school / bedroom / rooftop / rainy window
- Character types: quiet girl, gyaru, yamato nadeshiko, office lady, idol
Trained on my own illustrations. Free for personal use. Commercial use depends on the base model license (Anima is non-commercial).

