Back to Analyzer
Powered by Gemini 3

How VERITAS Works

AI can fake pixels. It cannot fake Newton. We use the laws of physics to expose AI-generated videos.

Analysis Pipeline

1

Video Upload

User uploads suspicious video

2

Object Tracking

Gemini 3 identifies and tracks moving objects

3

Trajectory Extraction

Extract position data (x, y, t)

4

Physics Analysis

Apply 6 physics laws to verify motion

5

Pattern Matching

Compare against known fake signatures

6

Verdict

AUTHENTIC or SYNTHETIC with confidence score

The 6 Laws of VERITAS

Gravity

y = v₀t - ½gt²

AI videos often have objects falling 15-50% faster or slower than Earth's gravity (9.81 m/s²).

If g ≠ 9.81 m/s² → FAKE

Momentum

p₁ + p₂ = p₁' + p₂'

In collisions, AI fails to conserve momentum - objects gain or lose energy impossibly.

Momentum violated → FAKE

Shadows

θ₁ = θ₂ = θₙ

AI often renders multiple shadow directions, revealing multiple impossible light sources.

Shadow angles don't align → FAKE

Reflection

θᵢ = θᵣ

Mirror and water reflections often don't match the actual object positions in AI videos.

Reflection mismatch → FAKE

Material

E = ½mv²

Glass that doesn't shatter, rubber that doesn't bounce - AI fails material physics.

Impossible material behavior → FAKE

Pendulum

T = 2π√(L/g)

AI pendulums swing at wrong speeds - the period doesn't match the length.

Period deviation → FAKE