Anime IRL (MiniMax H3)

A LoRA that creates anime-rendered subjects in a photorealistic world (this version only makes the subject's face/hair cel-shaded, while their body/limbs/clothes remain real). This is my first LoRA, so feedback is welcome :)

Trigger: ANIRL, anime-rendered face and hair

Strength: Start at 1.0, increase if the subject's face is too realistic

Example Prompt:

ANIRL, anime-rendered face and hair, a young woman with a blonde bob in a gray sweater sits at a kitchen table, slowly stirring a mug of coffee. Morning light comes through the window behind her. She looks down at the mug, then glances toward the window. Static camera, medium shot, natural lighting

Notes/Tips:

• Designed and tested for T2VA; no clue about I2VA

• Keep face/hair descriptions simple otherwise the model will push toward realism

• For multiple subjects, tag each one separately (e.g. an anime-rendered young woman, a fully photorealistic man)

• If faces are distorted, try a closer camera position. It doesn't do well with distant subjects

• I trained on mostly young adults, so older/mature subjects might require higher strength

Based on the awesome ANIRL Krea 2 style by @llikswonskcalb: https://civitai.com/models/2762978/bss-styles-for-krea2?modelVersionId=3113108