Home
Contact usLog inSign up
DEEPFAKE DETECTION

Detect AI generated calls

Expose AI-generated voice fraud in real time. Powered by Resemble Detect via Telnyx programmable voice.

GET STARTEDCONTACT US
Deepfake detection in the Telnyx Voice API
WHY DEEPFAKE DETECTION

Rebuild call trust that voice cloning broke

A cloned voice can authorize a wire transfer or impersonate an executive on a live call. AI voice fraud is projected to reach $415B in losses by 2028. Deepfake Detection catches synthetic voice at the network layer, on the same platform your calls already run on.

TALK TO AN EXPERT

  • Catch fraud mid-call

    Detection returns a real or fake verdict in under 4 seconds, fast enough to interrupt a fraudulent transaction as it happens, not just record one for the post-mortem.

  • Built into the network

    Detection runs natively on the Telnyx voice pipeline, not behind three vendor hops. No middleware, no third-party routing. Just enable it on your call.

  • Compliance-ready by design

    The EU Data Act and EU AI Act require transparency around AI-generated content, and US legislation is moving the same way. Detection gives you a documented audit trail for flagging synthetic voice.

FEATURES

Real-time protection, built on carrier-grade infrastructure

Built into the Telnyx Voice API, deepfake detection gives you real-time protection against AI-generated voice fraud. No extra infrastructure, no added latency.

  • Real-time frame-by-frame analysis

    Resemble Detect's neural model inspects audio frame by frame as the call happens, so even short inserted or altered segments are accurately flagged.

  • Verdict in under four seconds

    Detection returns a result fast enough to act on live. Interrupt the call, escalate, or route to a human before the fraud completes.

  • Voice API and TeXML

    Enable detection through the Voice API by adding the deepfake_detection object to your Dial or Answer command, or use it with TeXML, whichever fits your stack.

  • Webhook-driven results

    Receive a call.deepfake_detection.result webhook with a real / fake / silence_timeout classification, a 0.0-1.0 probability score, and a consistency percentage across frames, ready to wire into your fraud logic.

  • Zero call impact

    Detection runs in the background while the call proceeds normally. No impact on call audio quality or latency for either party.

  • Network-layer protection

    Detection runs at the network layer on the same platform your calls already use. No third-party routing, no signal lost across vendor hops.

DEEPFAKE DETECTION FOR DEVELOPERS

Enable detection in a single API call

Add the deepfake_detection object to your Dial or Answer command and handle the webhook. Full reference and code samples in the developer docs.

Explore Dev Docs

Enable on dial

POST https://api.telnyx.com/v2/calls

{
  "connection_id": "uuid-of-your-connection",
  "to": "+18005551234",
  "from": "+18005554321",
  "deepfake_detection": {
    "enabled": true,
    "timeout": 15,
    "rtp_timeout": 30
  },
  "webhook_url": "https://www.example.com/webhooks"
}

Enable on answer

POST https://api.telnyx.com/v2/calls/{call_control_id}/actions/answer

{
  "deepfake_detection": {
    "enabled": true,
    "timeout": 20,
    "rtp_timeout": 30
  }
}

Handle the result

{
  "record_type": "event",
  "event_type": "call.deepfake_detection.result",
  "payload": {
    "result": "fake",          // real | fake | silence_timeout
    "score": 0.97,             // 0.0-1.0 probability AI-generated
    "consistency": 92,         // % consistent across frames
    "call_control_id": "v3:..."
  }
}

Handle errors

{
  "record_type": "event",
  "event_type": "call.deepfake_detection.error",
  "payload": {
    "error_message": "detection_timeout",
    // detection_timeout | rtp_timeout |
    // dfd_connection_error | dfd_stream_error
    "call_control_id": "v3:..."
  }
}
RESOURCES

Everything you need to get started

From your first API call to handling the webhook, these guides walk you through it.

  • Icon Resources Article

    Developer documentation

    Full API reference, parameters, and code examples for enabling deepfake detection on the Voice API and TeXML.

    Learn more
  • Icon Resources Article

    Webhook reference

    Payload reference for detection results, including classifications, scoring, and signature verification.

    Learn more
  • Icon Resources Article

    Release note: Deepfake Detection

    Announcement and overview of deepfake detection with Resemble Detect.

    Learn more
PRICING

See our Deepfake Detection pricing

Deepfake detection is available to every Telnyx account. No setup fees, no minimums. Just add the deepfake_detection object to your Dial or Answer command, or enable it in TeXML.

SEE PRICING
$0.01per invocation

Ready to detect deepfakes?

From a single API call to a real-time verdict, Telnyx gives you carrier-layer deepfake detection built into the network your calls already run on.

GET STARTEDCONTACT US

FAQ

A real-time service that analyzes call audio to determine whether a voice is a real human or AI-generated. Telnyx uses Resemble Detect, a neural model that inspects audio frame by frame to flag synthetic or modified speech.

A real-time service that analyzes call audio to determine whether a voice is a real human or AI-generated. Telnyx uses Resemble Detect, a neural model that inspects audio frame by frame to flag synthetic or modified speech.