LiveKit Agents gives developers direct access to the caller's audio frames before speech-to-text runs. That's the right spot for a deepfake check. Forward the same frames to Resemble Detect's streaming API, and your agent gets a running result it can use to keep the conversation going, ask for verification, or hand off to a human.
Override the agent's audio input step to send caller audio to Detect as the session runs.
Use the result in your agent code to pause sensitive actions or hand off to a person.
Pair self-hosted LiveKit with an on-prem Detect deployment when audio has to stay on your servers.
Screen any participant's audio in LiveKit rooms, not just one-to-one agent calls.
