SceneDepth V2 — Cinematic Anime Composition & Spatial Depth LoRA

A composition-focused LoRA trained to create stronger foreground, midground, and background separation in anime-style illustrations.

SceneDepth V2 is designed for scenes where composition matters just as much as the character — museums, libraries, streets, interiors, stations, architectural spaces, and cinematic environments with clear visual depth.

✨ Best for
• Cinematic anime compositions
• Strong foreground framing
• Deep environmental perspective
• Layered foreground / midground / background scenes
• Architectural interiors
• Mature anime illustrations
• Off-center character placement
• Wide environmental shots
• Visual storytelling scenes

🎯 Recommended settings
LoRA Strength: 0.7–0.85
Recommended starting point: 0.8
Steps: 25–35
CFG: around 4

🏆 Recommended version
Step 300

SceneDepth works best when your prompt clearly describes the spatial relationship between foreground, subject, and background.

Example prompt concepts:
“frame the character through a large foreground object”
“strong foreground-to-background depth”
“layered environmental composition”
“distant architecture receding into the background”
“asymmetrical cinematic framing”

This LoRA is focused on composition and spatial depth, rather than changing the base model's character style.

The showcase images were generated using the actual trained LoRA.

More examples and updates coming soon.