SpeaRec is a speaker verification system

Verify the identity of your customers using voice biometrics and deep learning.

How it works 
// API request
/verification user="Bob" track="file.ogg"

// Returned object
{
  "verified": true,
  "score": 0.94
}

Content independent

Your customers will never be asked to use and remember a specific passphrase nor reveal personal information. Enroll and verify your customers' identity just by asking them to reply to any generic question.

Language agnostic

You can verify your customers worldwide. SpeaRec uses the latest AI technology to analyze voice biometrics independently from the language spoken by the customer.

Highly accurate

You can verify the identity of your customers by providing security that is more robust than most alternative verification services. Massive tests have shown that SpeaRec has an equal-error-rate (EER) lower than 0.001.

Secure by design

Voice biometrics can't be re-engineered into a voice stream. Audio tracks are never stored in a database. In the rare case of a data breach, your business and the privacy of your customers are absolutely safe.

Get started with SpeaRec

SpeaRec has been designed from the start for a simple and gradual adoption. You can use SpeaRec to provide an additional layer of security or as a stand-alone PSD2 compliant user verification service.

Get a taste of SpeaRec, the steps below will help you get started.

Sign Up

Sign-up into the Nextbit customer portal, you will gain access to a list of products based on artificial intelligence.

Get a SpeaRec API key

Within the SpeaRec product page, create one or more API keys.

The demo usage plane allows you to test SpeaRec without any commitment. Please contact us for details.

Enroll a user

With the /enrollment resource you can enroll a user. It will use the audio track to generate or improve the user voice signature. A new user needs at least three different tracks for being properly enrolled. 

// Call 3 times with 3 different tracks
/enrollment user="3141" track="file.ogg"

// Returned object
{
  "user": "3141",
  "number_of_enrollments": 1
}

Each of the audio files used during the enrollment must have a length of at least 7 seconds. At the end of the enrollment procedure, the audio files are deleted.

Verify user identity

With the /verification resource you can verify the user identity. It will use the audio track to establish if the voice belongs to the user or not. 

// Call once
/verification user="3141" track="file.ogg"

// Returned object
{
  "verified": true,
  "score": 0.94
}

The audio file will be automatically deleted at the end of the verification procedure.

Use voice biometrics to
>