Drop any image and get a forensic authenticity verdict in seconds. Detects AI-generated images and deepfakes — free, instant, and 100% private.
Drop an image here or click to choose a file
JPG · PNG · WebP — up to 25 MBNeed certainty? Pro Deep Scan runs ensemble neural models (GAN/diffusion fingerprints, face-warp detection) and covers video & voice — with a court-ready PDF report.
Upgrade to Pro →Every scan runs a battery of independent forensic tests, then fuses them into a single explainable verdict.
Scans embedded metadata, C2PA manifests and tool signatures left behind by Stable Diffusion, Midjourney, DALL·E, Firefly, Flux and 40+ other generators.
Re-compresses the image and maps compression error. Authentic photos show natural variation; synthetic or spliced regions glow with uniform error.
Real camera sensors leave organic, uneven noise. Diffusion models produce statistically smooth noise fields — we measure the difference block by block.
Checks resolution and aspect signatures typical of generator output grids (512/768/1024-class latents) and missing optical artifacts.
Verifies EXIF camera chains, software edit history and timestamp coherence — absence of provenance is evidence too.
Cloud ensemble of diffusion-artifact CNNs, face-landmark warp detectors, and voice-clone classifiers for video & audio.
Anywhere a fake face, fake voice or fake document costs real money.
Screen candidate photos and video interviews. Remote-hire deepfake fraud is now a top-3 vector for corporate infiltration.
Catch synthetic IDs and AI-generated selfies before onboarding. Batch API verdicts in under 300 ms.
Verify user-submitted imagery before publication, with an evidence trail you can cite.
Flag AI-faked product photos, fake reviews avatars and counterfeit listings at upload time.
Detect AI-doctored damage photos before paying out. One caught claim pays for a year of DEEF.
Romance-scam profiles are 80%+ AI-generated. Check before you trust.
Scan free in your browser. Buy deep-scan credits only when the stakes are high — pay per use, credits never expire.
Already purchased? Use your license key on this device.
Lost your key? Recover it by email
Drop DEEF into your KYC flow, moderation queue or claims pipeline in an afternoon. Verdicts return in under 300 ms with per-signal confidence breakdown.
# Detect a deepfake in one call curl -X POST https://api.deef.ai/v1/detect \ -H "Authorization: Bearer $DEEF_KEY" \ -F "file=@selfie.jpg" # → response { "verdict": "ai_generated", "risk": 0.94, "signals": { "diffusion_artifacts": 0.97, "sensor_noise": 0.91, "provenance": "absent" }, "latency_ms": 212 }
The free engine runs heuristic forensics (metadata, ELA, noise statistics, geometry priors). It catches the majority of unedited generator output and gives an explainable risk score, but it is screening-grade, not court-grade. Pro Deep Scan adds cloud neural ensembles for materially higher accuracy and covers video and voice.
Free scans never leave your browser — analysis is 100% client-side. Pro Deep Scans are processed in-memory and deleted within 30 minutes unless you save them to your history.
No, and anyone claiming otherwise is selling you certainty that doesn't exist. DEEF reports calibrated risk with per-signal evidence so you can make the judgment call — or escalate to our human forensics partners on Business plans.
Video frame-sampling and voice-clone detection are part of Pro Deep Scan. Per-minute API pricing is available for call-center and KYC pipelines.