Liveness Check Agent

Version: always use the latest available version.

The Liveness Check Agent is designed to ensure that the person behind the screen is a live person and is present during the process, without needing to leave the chat. It is recommended to use this agent when you only need to validate a person’s liveness. This agent can be used in any country.

This is not considered a complete biometric process.

Prerequisites

✅ Checklist to start without blockers

The user must be able to record video from the chat (camera permissions must be enabled on the device).
Recommended: define the handling flow if you enable Human in the loop (who reviews, response times, and what happens if there is no response).

What steps does the user complete?

Read instructions

The user sees a message in the chat with the steps to follow for recording a video and an image with instructions for a proper recording.

Record a video selfie

The user records a short video in which they must mention a unique numeric sequence (OTP).

Constant communication

The user receives messages indicating that their video is being analyzed: “We are verifying your video. This will only take a few seconds.” If the analysis is successful, they will see “Your liveness check was successful.” If the analysis encounters errors: “We had a problem analyzing the video. Please try the process again.”

How is the agent composed?

The agent contains tools that analyze the video submitted by the user:

OTP generation: generates the numeric sequence that the user must say when recording the video.
Speech to text: converts the audio to text to compare whether it matches the issued code.
OTP validation: validates that the mentioned code is correct.
Passive liveness validation: validates the liveness of the image extracted from the video submitted by the user. This process identifies and neutralizes deepfakes or any impersonation attempt, guaranteeing the presence of a live person’s face.
Lipsync validation (optional): validates that the video audio matches the video, comparing the audio frequency with the video frequency to confirm they belong to the same source.
Human in the Loop (enterprise only): if the process fails, you can opt for Human in the Loop (HIL), where the case will be reviewed by a human operator.

Connection from Marketplace

Access the platform

Go to apps.jelou.ai. On the home screen you will see Brain Studio and Connect. In Brain Studio, select Marketplace.

Find the agent in Marketplace

In Marketplace you will see three tabs: Catalog, Installed, and Developers. Click on Developers and use the search bar to find Liveness check agent. Then click the download icon to install it.

Confirm the installation

If the installation was completed successfully, you will see the Installed tag in green.

Open the agent in Brain

Go to Brain. In the sidebar you will see your installed skills and tools. Search for Liveness check agent and click to open it: the flow will load in your canvas.

Configure the version and inputs

Click on the input node, open Advanced settings (at the bottom), and select the agent version. If you will use default values, you do not need to modify inputs. If you are going to change any input or customize messages, review the Configuration section.

Configure the outputs

This agent has 5 error outputs and 1 success output. You can direct each output to:

A text input with a custom message.
Connect (only if your organization has this module).

Run tests

With the configuration ready, run tests using a skill or a prior agent as a “precursor” to trigger the flow.

Configuration

Inputs
Outputs

Language

string

default:"Es"

Variable: language. Defines the language of the experience.Available values: Es (Spanish), En (English)

Max. liveness check attempts

number

default:"3"

Variable: retries. Maximum number of liveness check attempts.Available values: 1-5

Notification email for exceeded retries

string

Variable: customerServiceEmail. Email to be notified when the maximum number of attempts is exceeded.

Enable Liveness Introduction Video

boolean

default:"false"

Variable: enableIntroVideo. Determines whether the introduction video should be shown to the user before starting the process.

URL to display liveness introduction media

string

Variable: introMediaUrl. The URL provided to the user to load an introduction video or image that will be used in the process.

OTP code digit count in Chat

OTP code length in Chat

number

default:"4"

Variable: otpLength. Number of digits in the OTP code.Available values: 3-6

Maximum OTP code duration

number

default:"1"

Variable: otpDuration. Maximum duration of the OTP code, in minutes.Available values: 1-10

Enable Human in the loop

boolean

default:"false"

Variable: enableHumanInLoop. This variable indicates whether the review process with a human agent should be activated.

Enable LipSync

boolean

default:"false"

Variable: enableLipSync. Enable LipSync for lip movement comparison with audio.

Max. agent attempts

string

default:"Unlimited"

Variable: retriesAgent. Maximum number of agent attempts before blocking the user.Available values: Unlimited, 1, 2, 3

User block duration

string

default:"1 day"

Variable: blockingInHours. Block duration when the agent attempt limit is exceeded.Available values: 1 day, 1 week, 1 month

Enable custom messages

boolean

default:"false"

Variable: enableCustomMessages. Enable custom messages.

Speech to text messages

Variable: speechToTextMessages. Custom messages for audio transcription errors.JSON format:

{
  "no_text": {
    "description": "When you recorded the video selfie we could not capture the sound of your voice. You may be in a very noisy place, or you sent a video without sound.",
    "advice": "Try again, making sure the video selfie has sound."
  },
  "error_http": {
    "description": "We're sorry, there was an error in the audio transcription.",
    "advice": "Try again, making sure the video selfie has sound."
  },
  "error_code": {
    "description": "We're sorry, there was an error in the audio transcription.",
    "advice": "Try again, making sure the video selfie has sound."
  }
}

OTP validation messages

Variable: OTPValidationMessages. Custom messages for OTP validation errors.JSON format:

{
  "client_error": {
    "description": "⚠️ The numeric sequence you mentioned is *[transcription]*, and it is not correct.",
    "advice": "Try again, making sure the numeric sequence is correct."
  },
  "OTP_ERROR": {
    "description": "⚠️ We're sorry, you exceeded the time limit for us to validate your video selfie.",
    "advice": "Try again, making sure the numeric sequence is correct."
  },
  "CODE_ERROR": {
    "description": "We're sorry, an unexpected error occurred in the service.",
    "advice": "Try again, making sure the numeric sequence is correct."
  },
  "HTTP_ERROR": {
    "description": "We're sorry, an unexpected error occurred in the service.",
    "advice": "Try again, making sure the numeric sequence is correct."
  }
}

Passive liveness messages

Variable: passiveLivenessMessages. Custom messages for passive liveness errors.JSON format:

{
  "resp_face": {
    "description": "We're sorry! We analyzed your video and *could not detect any face*.",
    "advice": "Try again, making sure the video selfie has a visible face."
  },
  "more_than_one_face": {
    "description": "We're sorry! We analyzed your video and *detected more than one face*.",
    "advice": "Try again, making sure the video selfie has only one visible face."
  },
  "face_far": {
    "description": "We're sorry! We analyzed your video and found that *your face was a little far from the camera*.",
    "advice": "Try again, making sure the video selfie has your face facing the camera."
  },
  "face_close": {
    "description": "We're sorry! We analyzed your video and it appears *you moved your face too close to the camera and a complete image could not be captured*.",
    "advice": "Try again, making sure the video selfie has your face facing the camera."
  },
  "face_cut_covered": {
    "description": "We're sorry! We analyzed your video and *it seems your face was covered or you were not well centered to the camera*.",
    "advice": "Try again, making sure the video selfie has your face facing the camera."
  },
  "eyes_closed": {
    "description": "We're sorry! We analyzed your video and could not detect your eyes well, it seems *you were wearing glasses or your eyes were slightly closed*.",
    "advice": "Try again, making sure the video selfie has your eyes open."
  },
  "low_quality": {
    "description": "We're sorry! The video we received is low quality and we could not process it correctly.",
    "advice": "Try again, making sure the video selfie has adequate quality."
  },
  "code_error": {
    "description": "We're sorry, an unexpected error occurred in the service.",
    "advice": "Try again, making sure the video selfie has adequate quality."
  },
  "error_http": {
    "description": "We're sorry, an unexpected error occurred in the service.",
    "advice": "Try again, making sure the video selfie has adequate quality."
  },
  "error_general": {
    "description": "We're sorry, an unexpected error occurred in the service.",
    "advice": "Try again, making sure the video selfie has adequate quality."
  }
}

Enable instructions

boolean

default:"true"

Variable: enableInstructions. Instructions in an image are enabled by default.

✅ Success

Liveness check successful

The liveness check was completed successfully.Variable: livenessSuccess

To access the information in this variable use the command $memory.get("response_liveness_success")

Response example:

{
  "response_liveness_success": {
    "url_video_liveness": "URL of the video sent by the user",
    "url_selfie_image": "URL of the frame obtained from the video to validate the user's liveness"
  }
}

Important fields:

url_video_liveness: URL of the video selfie submitted by the user
url_selfie_image: URL of the image extracted from the video to validate liveness

Information saved in Memory:The agent automatically saves the following information in memory variables:

{
  "livenessSuccess": true,
  "response_liveness_success": {
    "url_video_liveness": "https://cdn.jelou.ai/whatsapp_...",
    "url_selfie_image": "https://cdn.jelou.ai/video-fra..."
  }
}

Available variables:

$memory.livenessSuccess: Indicates whether the liveness check was successful (true) or not (false)
$memory.response_liveness_success: Object with the URLs of the video and the extracted image

You can access this information at any point in the flow using $memory.get("livenessSuccess") or $memory.get("response_liveness_success").

🔴 Errors

Time Expired

The time to complete the process has expired.Variable: timeExpiredError

Service error

An error occurred in the process of detecting the user’s liveness.Variable: serviceError

Process Abandoned

The user abandoned the process.Variable: userExitProcess

Liveness check unsuccessful

The user’s liveness could not be detected.Variable: livenessError

User Blocked

User Blocked.Variable: userBlocked

Frequently asked questions

Is this agent considered a biometric process?

No, the liveness check process is only a “stage” of the biometric process, therefore it is not considered facial biometrics.

What documents are required during the liveness check process interaction?

None. The only input is the video selfie that the user records with their phone.

Can I enable an instruction video?

Yes, you can enable the instruction video in the input Enable Liveness Introduction Video (true). Instructions in an image are enabled by default.

Can I add my own instruction media?

Yes, you can add your own instruction link in the input URL to display liveness introduction media, this way the default image or video is disabled.

If the user started their biometric process but has not finished it, how long before the process closes? Can the user resume the biometric process from the point where they left off?

The process closes when 1 hour passes without activity in the chat. If the user returns to the chat, they must start the flow from the beginning.

Getting Started

AI Agents

Brain

Nodes

Databases

Variables

Tools

Functions

Configuration

Marketplace

Enterprise

Billing

Prerequisites

✅ Checklist to start without blockers

What steps does the user complete?

How is the agent composed?

Connection from Marketplace

Configuration

✅ Success

🔴 Errors

Frequently asked questions

Getting Started

AI Agents

Brain

Nodes

Databases

Variables

Tools

Functions

Configuration

Marketplace

Enterprise

Billing

​Prerequisites

​✅ Checklist to start without blockers

​What steps does the user complete?

​How is the agent composed?

​Connection from Marketplace

​Configuration

​✅ Success

​🔴 Errors

​Frequently asked questions

Prerequisites

✅ Checklist to start without blockers

What steps does the user complete?

How is the agent composed?

Connection from Marketplace

Configuration

✅ Success

🔴 Errors

Frequently asked questions