<h1 id="overview">Overview</h1>A character LoRA trained natively on Anima Base v1.0 (Community DiT) capturing the facial structure, eye geometry, and styling of Kanna across multiple framing angles. Built with high concept disentanglement to keep facial identity locked while allowing full flexibility over outfits, poses, lighting, and backgrounds.### Usage Settings* Trigger Word: <code>kanna character</code>* Base Architecture: Community DiT (Kirazuri, Q-ANIMA, Z-Image, Anima Base v1.0)* Optimal LoRA Weight: <code>0.80 – 0.90</code> * <code>0.75 – 0.85</code>: Flexible styling, highly receptive to radical costume and background changes. * <code>0.85 – 0.95</code>: Maximum facial identity locking.* Sampler: <code>Euler_a</code>* Sampling Steps: 28 – 34* CFG / Guidance Scale: 4.0 – 5.0* Clip Skip: 1 or 2* Target Aspect Ratios: <code>832 × 1216</code> (Portrait) or <code>1024 × 1024</code> (Square)---* Prompt: masterpiece, best quality, 1girl, solo, upper body, detailed expressive eyes, distinct facial features, soft natural lighting, depth of field, looking at viewer* Negative Prompt:worst quality, low quality, score_1, score_2, score_3, bad anatomy, bad hands, missing fingers, extra digits, deformed limbs, disfigured, blurry, duplicate, cropped, watermark, text, signature, horns

Files with SHA256: 737deb8cd12041cc472ba3c55ecc8a8533979df4a9241177c18aa6af653b8a95