Comparing a face 1 to 1 answers one question: is it the same person in two images or not. Upload a photo from a document and a selfie, two documents or a profile picture — the platform finds the faces, compares them and returns the answer “same” or “different”. In the personal account without development, via API or in a KYC‑scenario. The decision on the answer is yours.
Comparing two images
The price follows the price matrix on the publication date
A passport photo and a selfie today: the platform finds the faces and compares them. “Different” on an old photo is a reason for manual review, not necessarily a rejection.
1:1 comparison, or Face Match, answers one question: do two faces match. The input is two images (a passport photo and a selfie, two documents, a camera frame and a scan); the output is “same person” or “different”.
Photos from documents, selfies, video frames and archive images are compared. One thing matters: a face of the required size must be visible in each image.
The platform returns the answer “same” or “different”, and in the personal account a confidence score next to it. Whether to accept, retake, send to an employee or reject is up to you: stricter for issuing money, softer for checking a profile.
In 1:1 comparison a face is matched against a single image. If you need to find out whether a person has appeared in your database, that is a face search 1:N. Repeat verification of a customer at login and payment is covered on the page about confirming operations by face.
The mechanics are the same — two faces are compared with each other. What differs is where the images come from and what the company does with the answer.
The customer has sent several documents — for example, a passport and a licence for a rental, or an old and a new passport after replacement. The comparison shows that all the documents belong to the same person, and not to different people within one application. Typical cases: car rental and car sharing, passport replacement at 20 and 45, an application with several documents.
On contractor marketplaces, dating and rental services, users upload a profile photo. Comparing it with a selfie taken live (Liveness) helps to understand that the profile shows the account owner and not someone else’s picture. Typical cases: contractor and seller profiles, a landlord’s profile photo, a change of profile photo.
An anti-fraud or security employee uploads two files to the “Face comparison” tab and clicks “Compare”. The answer “Same” — yes or no — and the confidence score are visible immediately; the decision stays with the employee. Typical cases: disputed applications and complaints, review of archived cases, handling customer requests.
At a pickup point, rental or office, an employee photographs the customer on a tablet and scans the document. The platform compares the faces, and the employee makes the decision — with no “looks about right” mistakes. Typical cases: handing over orders and valuables, equipment rental, the office of a microfinance company or pawnshop.
Four steps from two files to an answer. The general face recognition scheme is on the page about face recognition; here is what matters specifically for comparing two images.
In the document and in the selfie the platform finds the face and aligns it by landmarks. If the face is too small, cropped or turned too far, the answer will be “could not compare”, not a false “different”.
Head rotation, lighting, glasses and a mask are assessed. At the KYC step this works together with Liveness: it does not accept a screen capture, a printout or a mask before the faces are compared.
Each face is turned into a numeric vector — a biometric template. Templates are compared, not pixels.
The platform compares the two templates and returns the answer: “same” or “different”. In the personal account you see it immediately; via API you receive it in the response to the request and decide what to do next: accept, retake, send to an employee.
A passport photo may have been taken 10–20 years ago: a passport is replaced at 20 and 45. The larger the age gap, the higher the risk of a false rejection — for old documents add a repeat image and manual review of disputed pairs.
A hairstyle, a beard, glasses and a medical mask change the face in the image. The platform detects glasses, a mask, a beard and head position: if the face is covered or the frame is poor, it is better to retake than to get a random answer.
A three-quarter face, a shadow across half the face and poor lighting interfere with comparison. The platform detects head rotation and lack of light and asks for the frame to be retaken — this is not “different”, but “the image was unsuitable”.
A passport photo is small, laminated, with glare and a security mesh. So it is better to photograph the spread without glare, and the face in the scan must be no smaller than the set size.
Relatives and twins are the hardest case for any algorithm. When the cost of a mistake is high, a disputed match is sent for manual review.
The result depends on your document photos, cameras and lighting. The most reliable way is to test comparison in a test scenario with your own data. How face recognition works as a whole is described on the page about face recognition.
Most often “different” means not a fraudster but an unsuccessful pair of images. Here is what affects the result and what to do about it.
How recognition errors affect a KYC check and how to set accuracy requirements is covered in the article “Face recognition accuracy in KYC”.
Any comparison has two possible errors. There is no perfect rule — there is one that suits your risk.
The system said “same”, although the images show different people. This is a direct route for a fraudster with someone else’s document. So where an error is costly, the answer is supplemented by a Liveness check and an employee.
The system said “different”, although it is one person — for example, with an old passport photo. An honest customer leaves or writes to support. So “different” on an old document is a reason for a repeat image or manual review.
Issuing a loan, withdrawing funds, handing over valuables: here missing a substitution costs more than asking the customer to retake the frame. Any doubt goes to an employee.
Checking a photo in a profile, a repeat image in an app: here it matters more not to lose the user. The first “different” is a reason to retake, not to reject.
A photo on a screen will match the passport
A printout, a picture of a monitor or a video of a person gives the same face as in the document. Liveness does not accept an image in place of a live person before the faces are compared.
Deepfake and face swapping
A generated face or a face swap can resemble the photo in the document. So an online check uses detection of deepfakes and face swapping — more about it on the page about deepfake protection.
A live face is the reference point
In a KYC‑scenario with a live face, all other images are compared with the Liveness face: the photo in the document and a selfie with the document in hand. If the face in Liveness and in the selfie differ, the session is immediately flagged as suspicious.
When Liveness is not needed
If neither image comes from the user — for example, two scans of documents from an archive or photos an employee took personally at the counter — comparison is enough. Liveness is needed where the image is sent by the person you are checking.
Comparison tells whether two faces look alike. But if a fraudster shows the camera a photo of the passport holder, the faces will match perfectly. So in an online check, comparison always goes together with a live presence check — for example, with passive Liveness.
A face image used to establish whether it is the same person is biometric personal data. Separate rules apply to its processing by organisations in Russia. Below are the provisions your scenario depends on, as of October 2026. This is guidance, not legal advice: agree the scheme with your lawyer.
Comparison is needed wherever a document or an account may end up in the wrong hands.
In KYC, comparing the face with the document is included in the session price. Comparison via Face API and the “Face comparison” tab in the personal account is calculated for your volume.
The “Document + Liveness” package: document recognition and anti-fraud, a selfie, Liveness, comparing the face in the document with the selfie, detection of deepfakes and face swapping.
on request
Comparison of two images from your system: the answer “same” or “different” in the response to the request. Payment per request.
on request
The “Face comparison” tab for manual review of disputed pairs without development: upload two files — get an answer. Terms are available from the manager.
At 10 000 KYC‑sessions a year the “Document + liveness” package costs 33 ₽ per session, at 50 000 — 30 ₽, at 100 000 — 27 ₽. You pay only for completed checks; test scenario terms are available from a manager. For an exact calculation, see the calculator in the price matrix.
The scenario and access are configured in the personal account. Get a test scenario
A test scenario on your account, payment only for completed checks.
Terms you will meet in the personal account, in the API response and in conversation with a vendor.
The project «Development of a multimodal system using artificial intelligence "Neuromed" for managing the diagnostic processes of a clinic on the basis of the "Neurovision" computer vision platform» is supported by the Innovation Promotion Foundation within the federal project «Domestic Solutions»
The Neurovision computer vision platform is registered with Rospatent (certificate No. 2023661508) and included in the Russian Software Registry (entry No. 19972 of 13.11.2023)
The right to use the Platform is granted under a licence agreement concluded by Neurovision with its customers
Tell us which images you want to compare and where — in onboarding, in a user profile or at the counter — and we will suggest a way to connect and what to arrange with consent.
Can't put your request into words? Book a call and our team will contact you at a convenient time
Book a callThe company is included in the register of IT companies and the register of personal data operators of the Russian Ministry of Digital Development.
Address: Moscow, Presnenskaya Embankment 12 — Moscow City business centre, Federation Tower. Nearest metro stations: Delovoy Tsentr and Mezhdunarodnaya.
How face recognition works, how algorithm accuracy is tested and where the boundary of the law lies — in our blog
It is a comparison of two face images: whether the same person appears in them or not. It is also called Face Match or face verification. The platform finds the faces in both images, compares their biometric templates and returns the answer “same” or “different”.
In 1:1 comparison a face is matched against a single image: “is this the same person?”. In 1:N search, against all the faces in a database: “have we seen this person before?”. For searching your own database there is a separate feature.
A photo in a document, a selfie, a camera frame, a profile picture, an archive photo. The personal account accepts PNG, JPG and JPEG files. Each image must show a face of sufficient size without strong head rotation.
The result depends on the pair of images: the age of the document photo, angle, light and scan quality. So “different” on an old document is a reason for a repeat image or manual review. Accuracy for your flow is best checked in a test scenario with your own data.
Yes, it can: a passport is replaced at 20 and 45, and a document photo may be old. With a large age gap between the images a false rejection is more likely, so for old documents add a repeat image or manual review.
Send two images in a REST request and get the comparison result. For capturing a face on a website and in an app there is the Web SDK (in an app, via WebView or the system browser). Without development, comparison is available in the “Face comparison” tab in the personal account.
The comparison itself only answers whether the faces look alike: a photo of the passport holder will match the passport. So in an online check, comparison works together with Liveness and deepfake detection, which tell a live person from a photo, video and mask.
A face image used to establish identity is biometric data: written consent is required (Part 1 of Art. 11 of 152‑FZ), and 572‑FZ restricts the processing of biometrics by organisations outside the EBS. Legality depends on the purpose and the procedure of the check — agree the scheme with a lawyer before launch.
Yes, a new check can be linked to the customer’s previous successful KYC‑session, and the face will be compared with the face from it. This scenario — for login and confirmation of actions — is described on the page about confirming operations by face.
In KYC, comparing a face with a document is included in the “Document + Liveness” package — from 29 ₽ per session. Comparison via Face API and the personal account tab is calculated for your volume; the terms of a test scenario are clarified by a manager.
