Deepfake Detection API

What the Deepfake Detection API Does

The FauxLens deepfake detection API accepts an image or video file (or a URL to one) via a single authenticated POST request and returns a JSON verdict within roughly three seconds. Under the hood it runs six independent forensic signals in parallel - GAN fingerprint detection, facial-geometry consistency analysis, PRNU sensor-noise fingerprinting, temporal frame-artifact analysis for video, frequency-domain spectral analysis, and metadata forensics - then fuses them into a single confidence-scored verdict with a per-signal evidence breakdown.

This is the programmatic equivalent of the FauxLens web deepfake detector. The web tool is for one-off human checks; the API is for automating deepfake detection inside your own product - screening profile photos at signup, scanning uploaded media before publication, flagging suspected synthetic candidates in video interviews, or triaging user reports at volume. Every call returns the same structured evidence a human analyst would see, so you can set your own confidence thresholds and route flagged content into review.

Endpoint, Authentication, and Rate Limits

Authentication uses API key bearer tokens passed in the Authorization header, scoped per account and optionally restricted to specific IP ranges. The default free developer tier supports 60 requests per minute and 10,000 requests per month; higher tiers scale to 600 requests per minute with burst capacity. Rate-limited responses include a Retry-After header.

Supported inputs: JPEG, PNG, and WebP images up to 25MB; MP4 and MOV video up to 100MB. Both multipart form-data uploads and a JSON body with an image_url or video_url field are supported. All traffic is encrypted with TLS 1.3, and no submitted content is stored after the verdict is returned - the API runs on the same zero-retention model as the web tool. An OpenAPI specification, a Postman collection, and Python, JavaScript, and Go samples are available to approved developers.

Sample JSON Response

Every deepfake detection API call returns a consistent JSON envelope: a top-level verdict ("Authentic", "Likely Manipulated", or "AI-Generated"), a confidence float between 0 and 1, the suspected generation or manipulation method, processing time in milliseconds, a content hash, and an evidence array with one entry per forensic signal. Each evidence entry names the signal (for example gan_fingerprint, facial_geometry, temporal_consistency), a status (flagged, warning, or passed), a human-readable label, and a plain-language explanation of what the signal found. Your integration can act on the top-level verdict for a simple allow/deny gate, or inspect the evidence array to build a richer review UI for your trust-and-safety team.

Who Uses a Deepfake Detection API

Dating and social platforms use the API to screen profile photos and uploaded media at signup and on report, catching AI-generated personas used in romance scams before they reach other users. HR and recruiting tools screen candidate video and profile images to flag suspected deepfake interviewees - a documented and growing fraud vector. Newsrooms and content platforms scan submitted imagery before publication. Insurance and claims systems triage submitted damage photos. Marketplaces verify listing images. In each case the value is the same: deepfake detection that runs automatically, at machine speed, on every piece of content instead of relying on a human to notice.