Lighting Physics Validation

This checks if all highlights and specular reflections point to the same light source(s). AI often creates highlights that suggest conflicting light positions.

The detector analyzes shadow edges, penumbra gradients, and contact shadows. Real shadows obey inverse-square law falloff; AI shadows often have unrealistic hard edges or wrong intensity gradients.

Real scenes have consistent color temperature from light sources. AI images often mix warm and cool lighting inconsistently, creating unnatural white balance variations.

Ambient occlusion is the darkening in corners and crevices where light is blocked. AI often forgets to add natural AO or applies it inconsistently.

Yes, night photos often have multiple artificial light sources. AI frequently creates impossible lighting in night scenes, making them easier to detect.

Complex studio lighting creates multiple light sources, but they still obey physics. Real multi-light setups are consistent; AI multi-light attempts often have contradictory physics.

By analyzing highlight positions on known 3D shapes (spheres, faces), light direction can be estimated within 15-20 degrees for typical images.

HDR processing can change lighting appearance but maintains physical consistency. Stylized photos may trigger false positives, which is why this method has 9% weight in ensemble.

Lighting validation catches errors that pattern-based detectors miss. AI can match statistical patterns while still violating physics, making this a valuable complementary method.

Some AI systems are incorporating physics-based rendering knowledge, which may reduce lighting errors. However, complete physical accuracy requires 3D understanding that current AI lacks.