Upload a video or audio file. Clarifake analyzes visual, temporal, and audio signals to estimate whether synthetic generation may be present.
Every upload goes through the same forensic pipeline, whether it's a 10-second clip or a full interview.
Drop in a video or audio file, or paste a link. Clarifake extracts frames, waveforms, and metadata automatically.
Multiple independent models check for generation artifacts — spectral noise, lip-sync drift, unnatural blinking, and more.
Signals are combined into a single 0–100 score and a clear verdict: authentic, synthetic, or inconclusive.
You're not getting a black-box guess. Every score is built from four independent checks, and each one is shown to you separately.
Frequency patterns inconsistent with natural camera capture.
Frame-to-frame flicker, warping, and motion drift.
Frame 001 → 002 → 003 → 004 → ⚠ driftMismatches between speech and lip or facial movement.
Encoding history, C2PA credentials, and file provenance.
A score above 85 is flagged as high-confidence synthetic. Between 40–85, Clarifake returns "inconclusive" rather than guess.
Two clips of the same scene — one untouched, one AI-generated. The frame analysis is what tells them apart.
Verify footage before publishing.
Flag suspicious uploads automatically.
Preserve evidence and generate reports.
Check before you believe. Check before you share.
| Signal | Clarifake |
|---|---|
| Pixel artifacts | ✓ |
| Temporal consistency | ✓ |
| Audio-visual sync | ✓ |
| Metadata | ✓ |
| Confidence score | ✓ |
| Explainable result | ✓ |
Every scan ends with a verdict, a confidence score, and the underlying evidence — exportable as a forensic report you can attach to a case file or editorial review.
No account needed for your first three checks.