LoGG 2.5D

A rendering-focused model trained to improve lighting, material response and overall dimensionality while keeping the original anime aesthetic intact.

The goal was not to turn anime characters into photorealistic people, but to make the existing illustration feel more physically rendered — especially when it comes to fabrics, hair, skin, eyes and other surfaces.

### Training Details

- Base Model: HassakuXLIllustrious V22

- Training Images: 300 curated images

- Training Steps: 1,500

- Training Focus: Lighting, material rendering, depth and dimensional detail

- Trigger word: LoGG_25D

The dataset was curated specifically to teach the model how different materials react to light, with particular attention to fabrics and clothing. The result is a more dimensional 2.5D appearance with stronger surface definition and more convincing light/shadow interaction.

It works best when you want to retain the anime/illustration look while giving the image a more tactile and rendered appearance.

This is not intended to be a pure realism model or a conventional detail enhancer. The main idea is simple: make anime look more dimensional without making it stop looking like anime.

Recommended Use

Use it with your usual prompts and adjust the LoRA strength depending on how strong you want the rendering effect to be. Higher strengths can produce more pronounced material and lighting changes.

Have fun with it and feel free to experiment.