AntiDeepfake examines images and video with multiple independent forensic techniques, then returns a calibrated authenticity score and an exportable evidence report — through the dashboard or a single API call.
Built for the teams that can't afford to be fooled
Manipulated images and video are cheap to produce and increasingly difficult to spot by eye. Verification has to be systematic, documented, and repeatable.
No single technique is reliable on its own. AntiDeepfake runs complementary analyses in parallel and fuses them with a calibrated ensemble — so a borderline result is reported as inconclusive rather than a false accusation.
Detects face swaps and reenactment artifacts using a deep classifier applied to aligned facial regions.
FFT/DCT analysis surfaces the periodic grid patterns generative models leave behind in Fourier space.
Eye-state, micro-motion, and skin-texture cues separate live capture from screens, prints, and synthetics.
Optical-flow analysis across sampled frames flags flicker, warping, and unnatural motion in video.
Compares lip motion against speech energy and checks the audio spectrum for synthetic-voice signatures.
Attributes likely origin — GAN, diffusion, or genuine capture — from residual noise signatures.
Inspects EXIF, quantization tables, and editing-tool traces — and flags suspicious metadata stripping.
Error-level analysis, noise-consistency, clone detection, and lighting checks localize edited regions.
Upload through the dashboard, drag-and-drop in the browser extension, or POST to the REST API. Images and video supported.
Each forensic method runs independently and reports its own confidence; the ensemble fuses them into a calibrated score.
Receive an authenticity score, a per-method breakdown, localized anomaly overlays, and a hash-anchored evidence report.
A verdict is only useful if you can defend it. Every analysis is reproducible, attributable, and built to fit your data-governance requirements.
Each report includes the uploaded file's SHA-256 hash so the analyzed evidence can be verified later.
Run the full engine inside your own infrastructure with Docker — uploaded media never has to leave your network.
Every scan is logged with a timestamp, requester, and reproducible result record for audit review.
Set how long media and results are kept, or delete on demand to meet GDPR and internal policy.
Most verification vendors charge per check or thousands per month and hand you a black-box score. AntiDeepfake gives you the full breakdown and a one-time license.
| Capability | AntiDeepfake | Typical KYC provider | Enterprise detection SaaS |
|---|---|---|---|
| Pricing model | One-time license | $1–3 per check | $2,000+/month |
| Independent methods | 8 | 1–2 | 3–5 |
| Image + video | Yes | Images only | Yes |
| Per-method transparency | Full breakdown | Black box | Partial |
| Hash-anchored evidence report | Included | No | Add-on |
| Self-hosting | Supported | No | Rarely |
| REST API + extension | Included | Extra | API only |
Verify identity documents and selfie video to stop synthetic-identity fraud at signup.
Authenticate user-submitted media before publication and document the basis for the call.
Screen profile photos and uploads at scale through the API to keep platforms clean.
Produce hash-anchored forensic reports for claims, disputes, and evidentiary review.
Confirm that candidate photos and remote interviews are genuine, not generated.
Prove your own media is unaltered, or detect generated submissions in a pipeline.
Own the platform outright, or talk to us about volume, SSO, and support for larger deployments.
Secure checkout via Digistore24 · 60-day money-back guarantee
Tailored to your compliance and scale requirements
Because AntiDeepfake combines eight independent core methods and calibrates their fusion, it is more robust than any single model. No detector is perfect, so the platform reports a confidence level and an inconclusive band for borderline cases instead of forcing a real/fake label. You always see the full per-method breakdown behind a verdict.
Images (JPEG, PNG, WebP, BMP) and video (MP4, MOV, AVI, MKV, WebM). Video is analyzed using frame sampling for fast turnaround.
Yes. Media is processed for analysis only and is never sold or shared. Retention is configurable, and you can self-host the entire engine with the included Docker setup so data never leaves your environment.
Yes. Every license includes a personal API key and full REST documentation, so you can embed verification directly into your onboarding, moderation, or claims workflows.
The Professional license is a one-time purchase that includes the dashboard, REST API, browser extension, and self-hosting — with no recurring or per-scan fees.
Purchases are covered by a 60-day money-back guarantee through Digistore24. If it isn't a fit, request a full refund within 60 days.
Start analyzing media in minutes — or talk to us about an enterprise deployment.