Sealed RoseSealed Rose
RosterNot AI · Real
For creators →Sign in
📹 Real-Time Video Deception Defense

FaceTime Deepfake Challenge

Scammers use OBS virtual cameras and real-time face-swapping to fake live video calls. Use these 3 physical occlusion challenges to break synthetic filters live on camera.

📹
How Scammers Fake Video Calls

Scammers use software like OBS Virtual Camera and DeepFaceLive to inject real-time face swaps into FaceTime, WhatsApp, and Telegram. However, these models cannot handle physical 3D occlusions. Ask your match to perform these 3 challenges live on camera.

Current 3-Step Verification Protocol

Ask your match to perform these exact actions while on a live video stream.

01Hand Occlusion Test
Slow Hand Sweep Across Facial Features
Say This To Your Match:
"Please wave your open hand slowly directly in front of your mouth and nose for 3 seconds."
Why Deepfakes Fail This:

Real-time face-swap neural networks (like DeepFaceLive) rely on continuous facial landmark tracking. When fingers pass between the camera and face, the neural boundary is occluded, causing the synthetic face to detach, flicker, or warp onto the hand.

⚠️ Failure Signs To Watch For:

Noticeable glitching, face texture projecting onto fingers, or the face snapping back to the real person underneath.

02Extreme Yaw & Pitch Profile
90-Degree Profile Turn + Upward Tilt
Say This To Your Match:
"Please turn your head completely sideways to your left profile and look up toward the ceiling."
Why Deepfakes Fail This:

Most face-swap models are trained primarily on frontal and semi-profile datasets. When a subject turns past 75 degrees, the model is forced to hallucinate missing 3D contours, creating severe warping along the cheek and jawline.

⚠️ Failure Signs To Watch For:

Sharp boundary seam visible around the jawline, temporal blurring, or the face abruptly disappearing.

03Physical Object Contact
Hold a Solid Object Touching the Cheek
Say This To Your Match:
"Please hold a pen, spoon, or credit card directly against your cheek while speaking."
Why Deepfakes Fail This:

Real-time diffusion and GAN models cannot resolve foreign rigid objects intersecting with facial skin meshes. The AI tries to render skin textures over the physical object.

⚠️ Failure Signs To Watch For:

The object blends into the cheek like liquid, or the face clips through the item unnaturally.

Did they send a pre-recorded video clip instead?

If your match refuses a live call and sends a short video proof clip or selfie video, upload it to our free AI deepfake video scanner for frame-by-frame forensic analysis.

Check Video Clip →

Frequently Asked Questions

Can scammers fake a live FaceTime or WhatsApp video call?
Yes. By routing pre-recorded video loops or real-time face-swap software (like DeepFaceLive) through virtual webcam drivers (such as OBS Virtual Camera), scammers can stream an artificial face directly into mobile video calling apps on emulators or rooted devices.
Why do physical hand occlusions break real-time deepfakes?
Neural face-swap models rely on continuous, uninterrupted facial landmark triangulation (detecting 68 to 468 2D keypoints on the eyes, nose, and lips). When fingers or foreign objects pass across the mouth and nose, the neural network loses tracking, causing the fake face to distort, flicker, or project onto the hand.
What should I do if a match refuses to perform these simple gestures?
If an online contact constantly gives excuses for why they cannot turn their head, wave their hand in front of their face, or turn off a room light ("my camera is glitching", "bad connection", "I feel uncomfortable"), they are almost certainly using a virtual camera loop or deepfake filter.
How does Sealed Rose ensure video calls are 100% genuine?
Every creator on Sealed Rose undergoes mandatory biometric in-person liveness verification before onboarding. When engaging in live calls on Sealed Rose, you are communicating directly with an authenticated human creator with zero automated bots or agency intermediaries.