Photo to 3D Model#

Two-step pipeline that converts any photo into a 3D model:

1. Gemini preprocess — Transform photo into a clean white-background 45° isometric view
2. Tripo3D generate — Convert the isometric image into a .glb 3D model

Requirements#

• GEMINI_API_KEY — Get from https://aistudio.google.com/apikey
• TRIPO_API_KEY — Get from https://platform.tripo3d.ai/

Usage#

bash

# Full pipeline: photo → isometric view → 3D model
python3 scripts/photo_to_3d.py <image_path>

# Custom prompt for Gemini preprocessing
python3 scripts/photo_to_3d.py <image_path> --prompt "your custom prompt"

# Skip preprocessing (image is already a clean isometric view)
python3 scripts/photo_to_3d.py <image_path> --skip-preprocess

# Custom output directory
python3 scripts/photo_to_3d.py <image_path> --output-dir ./my_output

Output#

• output/{name}_isometric.png — Gemini-generated isometric view
• output/{name}_model.glb — Final 3D model file

Default Gemini Prompt#

The built-in prompt generates a 45° isometric "3D-printed model" style render with PBR materials on a pure white background. Override with --prompt for custom styles (e.g., game assets, architectural models, product renders).

Notes#

• Supported input: .jpg, .jpeg, .png, .webp
• Tripo3D generation takes ~1-3 minutes depending on complexity
• For higher quality, preprocess source images with upscayl before running
• Output .glb files can be converted to .usdz for Apple AR using usdzconvert