Hokkaido

anime LoRA featuring D-shaped gradient eyes, no pupils, lash-heavy style. trained on a carefully curated 128-image dataset of high-quality anime illustrations built around a signature aesthetic: distinctive D-shaped irises with tri-tone vertical gradient coloring, no pupils, flat heavy upper lash bars, needle-flick outer lashes, and clusters of high-gloss circular catchlights.

The dataset spans a wide range of characters, skin tones, hair colors, and outfit styles — from armored knights and cyberpunk kunoichi to Mediterranean guitarists and jungle botanists — ensuring the LoRA generalizes well across scene types rather than overfitting to a single character archetype.

Lighting and atmosphere were treated as first-class training signals throughout, with intentional variety across golden hour, moonlight, neon, candlelight, underwater caustics, snowstorm, and broadcast studio environments