The trust layer for the age of deepfakes
Know there's a real human behind every session
Deepfakes, AI-generated faces, and autonomous agents are getting past your signup flows, interviews, and support lines. Validia runs a 10-second camera check, fuses liveness and deepfake signals, and hands your backend a signed verified-human token you can trust.
Works in any modern browser · No app install · No PII in tokens
lvl: "basic" · exp: 10 min
~10s
median check time for users
3
live detection signals per session
1
API call to verify a token
0
PII stored in verified-human tokens
One check, a full stack of detection signals
A single liveness check is easy to fool. Validia layers independent engines — active liveness, deepfake forensics, and generative-AI detection — and fails closed if any of them says the session isn't a live human.
Active face liveness
LiveChallenge-response engine
A short face-in-frame challenge with randomized light sequences proves a live person is physically in front of the camera — not a photo, screen replay, or injected feed.
Deepfake detection
LiveDeepfake forensics model
Every session is screened against a purpose-built deepfake classifier that flags face swaps, reenactment, and synthetic face manipulation.
AI-generated media detection
LiveGenAI detection model
Separately scores whether the media was produced by a generative model (Stable Diffusion, Midjourney, GAN-family and similar), catching fully synthetic faces that aren't classic deepfakes.
On-device challenge fallback
LiveOn-device face landmarks
Randomized blink, smile, and head-turn challenges verified with on-device face landmarks — a lightweight engine that keeps verification working in constrained environments.
Multi-signal fusion scoring
LiveWeighted confidence engine
Liveness and deepfake signals are fused into one confidence score with a hard veto: if the liveness engine says it's not a live human, no token is issued regardless of the other signals.
Verified-human token
LiveSigned short-lived JWT
A pass mints a signed, short-lived token carrying the session, score, and assurance level. Your backend verifies it with one API call — no PII changes hands.
rPPG pulse detection
On the roadmapRemote photoplethysmography
Detecting a genuine human pulse from subtle skin-color changes on camera — a signal no screen replay can fake.
Anti-replay & injection defense
On the roadmapStream integrity checks
Defense against virtual cameras, feed injection, and replayed sessions at the transport layer.
Voice-clone detection
On the roadmapAudio deepfake screening
Screening call audio for synthetic and cloned voices, extending proof-of-human to voice channels.
From camera to trusted token in four steps
01
Drop in the check
Embed the verification screen with our web component, or redirect to a hosted check. Works in any modern browser — no app install.
02
10-second camera challenge
The user completes a quick face liveness challenge. Frames are simultaneously screened for deepfakes and AI-generated media.
03
Signals are fused server-side
Liveness confidence and deepfake risk are combined into a single score. Anything below threshold — or vetoed by the liveness engine — fails closed.
04
You verify the token
Your backend confirms the signed verified-human token with one API call, then lets the session through. Tokens expire in minutes and carry no PII.
Built for every flow where "is this a real person?" matters
Add a verified-human gate anywhere fraud, bots, or synthetic media cost you money or trust.
Remote hiring & interviews
Stop proxy interviewees and deepfaked candidates. Verify the person on the video call is real and live before and during technical interviews.
Fintech & KYC onboarding
Add a liveness and deepfake gate in front of account opening and high-risk actions, cutting synthetic-identity fraud before it reaches your KYC stack.
Marketplaces & gig platforms
Verify sellers, drivers, and freelancers are real humans at signup and re-verify at sensitive moments — payouts, account changes, disputes.
Call centers & support
Defend against deepfake vishing and social-engineering of support agents by requiring a verified-human check before account recovery or wire approvals.
Dating & social platforms
Kill romance-scam bots and AI-generated profiles. Issue verified-human badges that users can trust without exposing anyone's identity.
Online exams & education
Confirm the enrolled student — a live human, not a bot or a stand-in's replayed video — is the one sitting the assessment.
Creator & UGC platforms
Distinguish real creators from AI-generated personas, and give brands confidence the face fronting a campaign belongs to a live human.
Bot defense for communities
A one-time, 10-second human check at signup is a stronger and fairer gate than CAPTCHAs that LLM agents now solve better than people.
An API, not a dashboard project
The entire backend integration is one endpoint: post the token you received, get back whether it's valid and how confident we are. Tokens are short-lived signed JWTs, so a stolen token is worthless within minutes.
- ✓Hosted verification page or embeddable web component
- ✓Signed JWT with session ID, score, and assurance level
- ✓Single verifyToken call from your backend — any language
- ✓Fails closed: no live human, no token
# Your backend, after the user completes the check
POST /verifyToken
{ "token": "eyJhbGciOiJIUzI1NiIs..." }
# Response
{
"valid": true,
"session": "sess_8f2k...",
"score": 0.96,
"level": "basic",
"expiresIn": 600
}Privacy by default
Verification video is processed to produce a score — never stored without explicit user consent.
No PII in tokens
Tokens carry a session ID, score, level, and expiry. Nothing that identifies the person crosses into your systems.
Short-lived by design
Every verified-human token expires in minutes, so replaying or hoarding tokens buys an attacker nothing.
Start free, scale on your terms
Integrate and demo on the free Developer tier. When you're ready for volume, talk to us — pricing is per check and drops with scale.
Developer
$0 forever
Everything you need to integrate and demo.
- ✓50 verification checks / month
- ✓All detection signals
- ✓Test & live API keys
- ✓Detailed reports & webhooks
- ✓Community support
Growth
Most popularCustom volume pricing
For products putting real traffic through the check.
- ✓Volume-based per-check pricing
- ✓Higher rate limits
- ✓Priority email support
- ✓Usage analytics exports
- ✓Uptime SLA
Enterprise
Custom annual
For fraud, trust & safety, and security teams.
- ✓Custom thresholds & risk policies
- ✓Dedicated support & onboarding
- ✓Security review & DPA
- ✓Custom data-retention terms
- ✓Roadmap input (voice, anti-replay)
Frequently asked questions
What exactly do you detect?+
Today: active face liveness via a challenge-response camera check, deepfake manipulation, and AI-generated media, plus an on-device challenge fallback. The signals are fused into one confidence score and delivered as a signed verified-human token.
Is this identity verification / KYC?+
No — and that's the point. We prove there's a real, live human in the session without collecting IDs or storing biometric identity. We sit in front of (or alongside) your KYC stack as a fraud gate, not a replacement for it.
What happens to the camera footage?+
Verification media is processed to produce the score and never stored without explicit consent. The token your systems receive carries only the session ID, score, assurance level, and expiry — no PII.
How do we integrate?+
Frontend: embed the check or redirect to a hosted verification page. Backend: verify the returned JWT with a single API call. Most teams are running end-to-end in under a day.
How long does a check take?+
About 10 seconds for the user, end to end, in any modern browser on desktop or mobile. No app install, no extension.
See it catch a fake in the next 60 seconds
Run the live check yourself — then try to beat it with a photo, a replay, or an AI-generated face.