Origin ยท 15% of the IQ Score

Understand the origin of every asset.

Origin is the public name for Originality: how likely it is that a piece of work is an authentic, human-made original rather than AI-generated or a near-duplicate of something that already exists, read from the pixels themselves.

What it reads

The visual content of the work alone. Origin never looks at the brief or the brand profile, only the frames or image, so its read stays independent of everything else the run touches.

Human-origin confidenceweight 0.50

The likelihood that the work came from authentic optical capture rather than a generative model, weighed most heavily of the three because it is the question most likely to carry regulatory weight.

Content uniquenessweight 0.30

Whether the work's visual fingerprint is distinct from known reference material, or a near-duplicate of something already in circulation.

Visual authenticityweight 0.20

Observable signs of real optical capture: sensor noise, depth-of-field physics, chromatic aberration, natural lighting inconsistency, the opposite of the hyper-consistency typical of synthesis.

Origin, as it shows up in a run
Origin95
โœ“Human-origin confidence
โœ“Content uniqueness
โœ“Visual authenticity

What it doesn't do

Returns confidence, not a verdict

A borderline read stays visibly borderline instead of getting rounded to a clean yes or no.

Evaluates the visual content directly

A file can claim anything in its EXIF data. Origin looks at the actual pixels, which is why it is a separate signal from Custody rather than folded into it.

Works from the asset itself

Visual authenticity and human-origin confidence work from the asset alone; uniqueness improves as the reference corpus grows, but the other two dimensions don't depend on it.

Get a real read on human-origin confidence

Human-made, synthetic, or somewhere between, scored from sensor noise, depth-of-field physics, and lighting consistency in the actual pixels.

Get early access