AI-generated voice fraud is growing in volume and sophistication. Telnyx now offers Deepfake Detection with Resemble Detect through the programmable voice, a state-of-the-art neural model that exposes deepfake audio in real-time, helping you detect and flag artificially generated or modified voice content.
Resemble Detect uses a sophisticated deep neural network that is trained to distinguish real audio from spoofed versions. It analyzes audio frame-by-frame, ensuring any amount of inserted or altered audio can be accurately detected.
The call proceeds normally while detection runs in the background. There is no impact on call audio or latency.
As AI-generated voice content becomes more common, the ability to verify what's real isn't optional. Deepfake Detection catches fraud at the network layer, on the same platform where your calls already happen.
Also, the EU Data Act requires transparency around AI-generated content and data usage. Deepfake Detection with Resemble Detect helps meet these requirements by providing real-time verification of audio authenticity, giving you a documented way to identify and flag synthetic voice content.
deepfake_detection object to your Dial or Answer command via the Voice API:Outbound call:
Inbound call:
Handling results:
You receive a call.deepfake_detection.result webhook with:
result — real (human voice), fake (AI-generated), or silence_timeout (no speech detected)score — Probability the audio is AI-generated (0.0–1.0)consistency — How consistently the model classified audio across frames (0–100%)If detection fails, you receive a call.deepfake_detection.error webhook with error types like detection_timeout, rtp_timeout, dfd_connection_error, or dfd_stream_error.
For the full API reference and code examples, see the Deepfake Detection docs.
Ready to protect your calls? At $0.01 per call, deepfake detection is available to every Telnyx account today.