TL;DR
This LoKr was made to draw delicious boobies and female anatomy — and it does. Unexpected bonus: it also enhances contrast and sheen across the board, improving metals, glass, liquids, fabric, and other reflective surfaces.

LoKr Overview

Inspired by a cultured hentai artist known for beautifully rendered female anatomy. The focus is on strong contrast, clean highlights and shadows, and luminous, glossy skin. If the skin looks polished and glowing, it’s doing its job.

Training Notes

Trained mostly in the low noise stage to maximize texture and detail learning, with some mid noise training to reduce composition bias. The goal was to learn surface quality, not poses or scene structure.

Use Case

Best suited for illustrative or painterly styles where skin sheen and rendering quality matter. Not recommended for anime faces — all anime faces in the dataset were mosaic blurred to avoid interfering with character LoRAs. (Note sample images were blurred in POST. Not from LoKr use)

Recommended Settings

• Strength: 1.3

• CFG: 4

• Positive prompt: Cinematic photorealistic dark fantasy oil painting, hyper-detailed textures. Creamy, lustrous texture on her skin while casting gentle, defining shadows. Dramatic warm–cold lighting contrast with volumetric atmosphere, and focused composition reminiscent of high-end concept art and movie posters.

• Negative prompt: pixelated, censorship, mosaic, blur

Other Notes

Contains some NSFW knowledge, but don’t rely on it heavily. Background generation should remain unaffected, as the dataset consisted primarily of nude bodies on plain backgrounds. Although dataset consists of all body types, it usually lands on the more voluptuous side, therefore I'd recommend using a body type slider to achieve a desired result.

Feedback is always appreciated.