Turn detection

Turn detection is the process of deciding when a user has finished speaking, so the avatar can respond at the right moment without interrupting or waiting too long.

How turn detection works

Turn detection is the system's judgement about when the user has finished speaking and the avatar should respond. It is more subtle than detecting sound; the system has to distinguish a pause from the end of a thought.

In real-time avatar products, turn detection usually combines voice activity detection, ASR signals and timing rules. If it fires too early, the avatar interrupts. If it fires too late, the conversation feels slow.

A concrete example: a user says, "I want to connect Salesforce... but only for new leads." Good turn detection waits through the pause and responds to the full request.

Turn detection is one of the reasons live agents feel natural or awkward. It sits right at the boundary between audio infrastructure and conversational behaviour.

What Anam ships

Anam's Cara-4 model delivers expressive real-time avatars with around 150 ms server-side avatar-generation latency once a session is running, across 70+ languages. Builders use JavaScript and Python SDKs or integrations for LiveKit, Pipecat, ElevenLabs Agents, Agora, and VideoSDK. Bring any AI stack including OpenAI, Claude, Gemini, Mistral, Groq, Deepgram, Cartesia, or custom providers. The platform supports WebRTC delivery, SOC 2 Type II, HIPAA, zero data retention, and regional data residency. Sessions stream low-latency audio and video to browsers and native apps.

Explore the real-time avatar API →

Related terms

Barge-in

Barge-in is the ability for a user to interrupt an avatar while it is speaking, so the system can stop the current response, listen to the new input, and continue the conversation naturally.

ASR (speech recognition)

ASR, or automatic speech recognition, is the system that converts live spoken audio into text so an avatar or agent can understand the user and decide what to do next.

AI voice agent

An AI voice agent is a speech-first conversational agent that uses speech recognition, an LLM or agent workflow, and text-to-speech to understand spoken input and reply naturally in real time.

Sub-second latency

Sub-second latency is the end-to-end response time under one second between a user speaking and an avatar replying, covering speech recognition, model reasoning, speech synthesis, animation, and network streaming.

Frequently asked questions

What is turn detection in an avatar conversation?

Turn detection decides when the user has started, paused, interrupted, or finished speaking, so the avatar knows when to listen and when to respond.

Why does turn detection matter for real-time avatars?

Poor turn detection makes the avatar interrupt too early, wait too long, or miss a barge-in. Good turn-taking is one of the fastest ways to make the interaction feel human.

Is turn detection the same as speech recognition?

No. Speech recognition converts audio into words. Turn detection focuses on timing, silence, interruption, and whether the conversational floor should move to the avatar.

What makes turn detection difficult?

Real users pause, restart, think out loud, speak over the avatar, and talk in noisy rooms. The system has to distinguish hesitation from a finished turn.

Last updated: 17th July 2026 · Reviewed quarterly.

← Back to glossary