AI has made fake photos, videos, and voices convincing and cheap — and that has made authenticity a live question in courtrooms, disputes, and reputations. Deepfake analysis examines a file for the signs of manipulation and delivers a supportable opinion on whether it is genuine.
How can you tell if media is manipulated?
- Metadata analysis — creation data, editing software traces, and gaps
- Visual artifacts — inconsistent lighting, shadows, warped edges, unnatural blinking
- Audio anomalies — splicing, unnatural cadence, and synthetic-voice signatures
- Compression and sensor forensics — error levels and camera fingerprint mismatches
- Provenance — whether the file is an original or a re-encoded, re-shared copy
What types of media can you examine?
Images, video, and audio — including face-swap and lip-sync deepfakes, AI-generated images and cloned voices, spliced or edited recordings, and doctored screenshots. We assess both whether content was fabricated and whether an ostensibly original file has been altered. ROHOVOT operates under a licensed California investigator (California BSIS PI License No. 190161) and a court-qualified computer-forensics expert witness.
Will the analysis hold up in court?
Yes — when it uses recognized methods, preserves the file with an unbroken chain of custody, and is explained by a qualified examiner. Our approach follows the media-authentication and forensic-science principles supported by the National Institute of Standards and Technology, so the opinion is defensible under challenge. See NIST’s media forensics work at nist.gov. Give us the original file, not a screenshot, whenever possible.