Gemini sketches forensic reconstructions from photos

What stands out here is not factual accuracy but workflow design: Gemini is being pushed into a vision-reasoning-image-generation loop that feels like a prototype for future 3D and forensic tooling.

• –The custom prompt turns Gemini into a mixed-media analyst that alternates between text reasoning and generated visual artifacts instead of stopping at plain description
• –The demo hints at real developer-adjacent uses in previsualization, site inspection, scene reconstruction, game asset planning, and rough architectural analysis
• –The limitations are just as important as the wow factor: commenters called out hallucinated details, weak geometric confidence, and the lack of scale-accurate outputs like usable DWG or 3D files
• –Because the behavior is prompt-driven rather than a native product mode, reproducibility is shaky and likely depends on model tier, safety settings, and access to Gemini image generation