Face Matching API for selfie-to-ID verification
Compare a customer selfie with the portrait on an identity document and return a configurable similarity decision for your KYC workflow.
Document
SelfieClients who trust us
Three reasons to match a face to the document
A stranger's passport is useless
A fraudster cannot pass verification with someone else's or a forged ID: the selfie will not match the document photo. Liveness prevents a photo from standing in for a live person.
One person — one account
Face Search (1:N) finds duplicate registrations across your database — even if the customer changed their passport or details.
A threshold that fits the scenario
The similarity threshold is configurable: strict for KYC, light for fast onboarding. The higher it is, the fewer impostors pass — and the more often a genuine customer is sent to a retry.
One person — one account, even with a new passport
Face Search looks for one face among many in your database. On every registration it checks the customer against the whole base, so a repeated attempt is rejected even when the passport or details changed.
Doesn't confuse different people and doesn't reject your own
The model returns a similarity percentage and you compare it with a threshold configured in the dashboard. A genuine customer clears it while someone else's face stays well below it.
Recognizes the customer, even when they've changed
A passport photo may have been taken years ago — people change. The algorithm focuses on biometric traits resilient to age, hairstyle, weight and image quality. And the reverse — it does not confuse similar-looking people where simpler systems fail.
One person — different looks
Similar — but different people
Face biometrics never leave your loop
Work in our cloud or deploy the Face Match core into a closed perimeter on your own servers (On-Premise) — selfies, document photos and biometric templates stay with you.
The result: biometric data does not leave the company perimeter — covering regulatory and security requirements.
Classic integration for developers
If your team prefers to build the architecture by hand — use our ready-made SDKs and API.
Web SDK / Flow Widget
A ready-made JS camera component (widget) for your website or embedded mobile browser (Webview). Fully customizable design to match your brand book.
Mobile SDK
Native libraries for mobile apps. Deep access to the smartphone camera and biometric micro-packet transfer even on weak 3G connections.
REST API
Direct access to face-to-document matching endpoints for your server-side logic. Authorization via a private X-API-KEY.
MCP server for AI agents
AI agents connect directly to verification — without writing code. Compatible with Claude, GPT and any MCP client.
What face matching returns
The response is JSON: the decision, the numeric similarity score and the reason a comparison failed. You set the threshold at which a match counts, based on the risk in your scenario.
Key fields
| Field | What it means |
|---|---|
| result | The comparison decision: the two faces match or they do not. |
| prediction | Similarity score from 0 to 1 — 0.9636, for example. |
| prediction_percent | The same value as a percentage, easier to show an operator. |
| failure_reason | Why the comparison could not be made. |
| backend_session_id | Session identifier for audit and for reviewing disputed cases. |
Common errors
| Error | What to do |
|---|---|
| No face found | There is no face in the photo — ask the user to retake it. |
| Not allowed file, valid extensions: .jpg .png .jpeg | Unsupported file format — convert it before sending. |
| Client does not have access to Face2Face technology | No active subscription for the technology, or access is switched off. |
| Face2Face result does not exist | No result yet: send the pair of images for comparison first. |
You can compare two uploaded photos, or images from sessions that already ran — liveness or document recognition — so the selfie and the document portrait do not have to be sent twice. See the Face2Face reference
Numbers and facts
- 433+
- document types —Passports and hundreds of other document types supported worldwide.
- 180+
- supported countries —Global coverage for checks across jurisdictions.
- 40 sec
- onboarding time —Customer registration and verification in under a minute.
- 99%
- technology accuracy —High-accuracy AI-powered face and document verification.
Questions about face matching
Next in the flow: liveness detection · recognise and verify the document · find a duplicate in your base (1:N) · what a check costs
Verify faces instead of taking documents at face value
Create a developer account, build a Face2Face scenario in the builder and test matching on your own flow — 500 checks a month free, with no card.