AI video interaction glossary
Full duplex means audio can travel in both directions at the same time. In an AI call, the microphone can capture your speech while the character is speaking. That makes overlapping conversation possible, but whether the AI stops and responds to an interruption is a separate behavior.
Definition
Full duplex describes a communication channel that can send and receive simultaneously. For an AI voice or video call, it means the system can receive your microphone audio while playing the character’s voice. It does not by itself promise natural turn-taking, an immediate interruption response or zero delay. Those depend on how the application detects speech and manages the current reply.
By the Prelulu editorial team · Updated
Defines simultaneous sending and receiving. This networking definition does not establish an AI app’s interruption quality or latency.
Implementation documentation for detecting speech, interrupting an agent and handling false interruptions.
A character starts describing a busy station. You say, “Wait, make it a quiet seaside town,” before the reply ends. A full-duplex channel can carry your words during that reply; an interruption handler decides whether to stop the old audio and change the scene.
A walkie-talkie-style interaction sends one side at a time. Full duplex removes that channel restriction. People may still choose to take turns, and an AI application may still wait for a pause before answering. The label describes simultaneous communication, not conversational skill.
Check whether the microphone remains active during playback, whether the character actually yields to an interruption, and how quickly the next reply follows. Background noise, echo control and connection quality can affect the experience even when the audio channel is full duplex.
FAQ
No. Full duplex allows simultaneous audio in both directions. Barge-in is the application behavior that recognizes your interruption and stops or changes the AI’s current response. They often work together but answer different questions.
No. Audio direction and camera input are separate capabilities. A call can be full duplex with your camera off, and an animated character on screen does not prove the AI receives your video.
No. Speech detection, model generation, audio playback and the network can still add delay. Test the actual interruption and response behavior rather than treating the term as a latency measurement.
Choose a character for conversation, roleplay or a live call. Check the current plans and available credits before calling.
Explore AI charactersUpdated
