Voice Conversations
Veloquence supports real-time voice conversations with sub-second latency. Visitors click the widget and start talking — no downloads, no phone calls.
Veloquence's voice mode delivers real-time, bi-directional voice conversations with ultra-low latency for a natural, phone-call-like experience.
Veloquence uses a WebRTC-based voice pipeline with three stages: speech recognition converts the visitor's voice to text in real time, a frontier language model processes the conversation and generates a response, and natural text-to-speech delivers the agent's reply in a human-sounding voice. End-to-end latency is under one second in standard conditions.
Pair your agent's voice with a lip-synced video avatar for a more engaging, face-to-face experience. Avatars render in real time, synchronizing mouth movements and expressions with the agent's speech. Browse available avatars in your agent settings.
Voice sessions are automatically recorded and transcribed. Recordings are available for playback and transcripts are accessible from the conversation history in your agent's analytics dashboard. Both can be used to review what was discussed, identify knowledge gaps, and improve your KB over time.