Deepfake defense for video KYC & banking onboarding
Check onboarding selfies, ID photos and video for face-swaps and AI generation.
From $15 for 500 credits · No contract · Credits never expire
curl -X POST https://sealedrose.com/api/v1/detect/video \
-H "Authorization: Bearer sr_live_..." \
-F "[email protected]"{
"aiGenerated": true,
"confidence": 0.994,
"detectedModelLabel": "FaceFusion / Deep Live Cam",
"checkId": "chk_8f1a2e3b4c5d",
"cost": 12,
"creditsRemaining": 488
}Key integration use cases
Selfie and ID document inspection
Verify applicant selfie and ID photos for AI generation and face-swaps before authorizing bank accounts.
Video KYC recording auditing
Run automated frame-by-frame deepfake analysis on completed onboarding video sessions to prevent synthetic account takeovers.
High-value wire video authorization
Screen recorded video confirmations for face-swaps and generative video impersonation before authorizing high-value transfers.
Forensic API endpoints
Single unified authentication token with predictable credit pricing.
| Endpoint | Input | Latency | Credits |
|---|---|---|---|
| POST /api/v1/detect/image | Photo (JPEG, PNG, WebP) | 500–800 ms | 1 credit |
| POST /api/v1/detect/audio | Audio (MP3, WAV, M4A) | 0.6–1.2 s | 1 credit |
| POST /api/v1/detect/video | Video (MP4, MOV, WebM) | 2.5–4.5 s | 12 credits |
| POST /api/v1/detect/text | Plain text / chat message | Not yet measured | 1 credit |
| POST /api/phone-lookup | JSON { "phone": "+1..." } | < 250 ms | 1 credit |
Frequently asked questions
How do fraudsters attack modern video KYC verification?
Fraud syndicates inject software virtual cameras (OBS Studio, ManyCam, Deep Live Cam) into the browser or emulator environment, streaming real-time deepfakes of stolen identities directly into the KYC selfie and liveness capture window.
Why do active liveness prompts (e.g. head turns up/down) get bypassed in video KYC?
Basic biometric SDKs only verify whether facial landmark coordinates moved as requested (e.g., turning head up or down). Generative face reenactment models can replicate these motions in real time. Sealed Rose looks beyond landmark geometry to analyze sensor noise patterns, frequency spectra, and ear/jaw boundary blending that expose synthetic injection.
Can Sealed Rose detect virtual camera injection?
We check the recorded selfie or video clip for face-swaps, diffusion artifacts, and temporal inconsistencies. Because we analyze the media pixels rather than the local OS driver, we catch synthetic artifacts regardless of the injection tool.
How does Sealed Rose handle customer media and verification data?
Media is processed strictly in memory and deleted after the scan completes. Verification hashes and forensic breakdowns are stored without retaining raw biometric templates.
Deploy deepfake defense in your platform
Prepaid API credits start at $15 for 500 checks. Instant provisioning with zero contract or monthly commitment.