Voice cloning has made it easy to sound like a customer, an executive, or a family member on the phone. Telnyx and Resemble AI built deepfake detection directly into the Telnyx Voice API, so teams don’t need to fork audio, run a media server, or build their own integration.
You turn detection on per call by adding one parameter to the Answer or Dial command. Telnyx streams the other party’s audio to Resemble Detect, and the result comes back as a webhook with a verdict, a score, and a confidence value. Your application decides what happens next: continue the call, add a verification step, or send it to a fraud queue.
Every Resemble voice generated through Telnyx is embedded with an imperceptible PerTh watermark at creation. The signature survives telephony compression and transcoding, enabling downstream verification of any audio's origin.
Resemble Detect analyzes inbound call audio frame-by-frame to identify AI-generated or cloned voices. Enable via the Telnyx dial or answer APIs and receive alerts at your webhook URL.
Resemble Detect provides real-time audio authenticity verification and documented synthetic voice flagging — supporting EU AI Act transparency requirements and HIPAA, GDPR, and SOC 2 Type II regulated deployments.
For regulated industries, Zero Retention Mode processes all audio entirely in-memory — no transcripts generated, no speech data stored. Acoustic signatures only.
Detection is billed per invocation on your Telnyx account, with no setup fees or minimums.
Use Resemble through the Telnyx Voice API for programmable call control, Voice AI Agents for managed conversational AI, or the Standalone TTS API for direct synthesis whichever fits your architecture.
