Top news
Investing in the public benefit OfficialAnthropic NewsOct 18Popular Musk Pledges SpaceX Shares to ‘Every Kid in America’ WebBloomberg TechnologyOct 19Popular Websites are closing their doors to AI agents WebThe Verge AIOct 17 This is the new ChatGPT Voice, powered by GPT Live OfficialOpenAI on YouTubeOct 18 How AI giants will reshape global philanthropy WebAxios AIOct 19 Gemini 4 Argon Really Surprised Me WebTwo Minute PapersOct 19 Google Gemini 4 agents just beat Minecraft with no human intervention WebEngadget AIOct 19 Who could it be? 🧐 OfficialGoogle on YouTubeOct 18 Anthropic gives outside AI watchdog “employee-like” access WebArs Technica AIOct 18 What are self-replicating ‘AI worms’, and why is Anthropic worried about them? WebThe Guardian AIOct 18
GlossaryProductsIntermediate

Full-duplex

Also written: Full-duplex architecture, Duplex

Full-duplex communication allows simultaneous two-way transmission of data, like a phone call where both people can talk and listen at once.

In plain terms

Full-duplex means that information can travel in both directions at the same time. Think about a phone call: you can talk while the other person is also talking, and you can both hear each other. This is different from a walkie-talkie, where one person has to finish speaking before the other can reply.

In AI, full-duplex voice models allow for much more natural and responsive conversations. Instead of waiting for the AI to finish its response before you can speak, you can interrupt or talk over it, making the interaction feel more like talking to another human.

The expert version

Full-duplex refers to a communication channel that permits simultaneous transmission of signals in both directions. In the context of AI, particularly voice interfaces, this enables a system to both receive user input (speech) and generate output (speech) concurrently. This contrasts with half-duplex systems, which can only transmit in one direction at a time, requiring a turn-taking mechanism.

Implementing full-duplex voice AI requires sophisticated handling of audio streams, including real-time transcription, speech synthesis, and potentially complex logic for managing overlapping speech and agent delegation. The benefit is a significantly more fluid and less frustrating user experience, closer to natural human conversation.

Why it matters now

Full-duplex is a key feature for advanced voice AI, allowing for more natural and responsive interactions. Recent developments, particularly from OpenAI with models like GPT-Live, highlight its importance in creating AI that can listen and speak simultaneously, mimicking human conversation.

Where it shows up 9 items, newest first
How each company uses it
OpenAI ChatGPT
OpenAI has prominently featured full-duplex capabilities in its GPT-Live models, enabling real-time, simultaneous listening and speaking for more natural voice conversations through its API.
Related terms

Real-time AISpeech recognitionSpeech synthesisConversational AILatencyHalf-duplex

Often confused with
Half-duplex
A communication system that allows transmission in only one directi
Echo cancellation
A technique used in telecommunications to remove unwanted ech
Simultaneous interpretation
While it involves real-time processing, this is abo
Origin

The term 'full-duplex' originates from telecommunications engineering, describing communication systems that allow simultaneous two-way data flow. It's a standard concept in networking and telephony. Source ↗

Try this about 5 minutes

Engage with a voice assistant that supports full-duplex communication. Try interrupting the assistant mid-sentence or speaking a question while it is still responding to your previous query. Notice how the system handles the overlapping audio and attempts to respond naturally, unlike older systems that would cut you off or ignore you.

Try it: Playbook entries that use thisPlaybook →

The Playbook: reusable how-tos, prompts, configs and tips pulled out of the news.

Questions people ask

What's the difference between full-duplex and half-duplex?
Full-duplex allows both parties to speak and listen at the same time, like a phone call. Half-duplex only allows one party to transmit at a time, like a walkie-talkie where you press a button to talk.
Why is full-duplex important for AI voice assistants?
It makes conversations feel more natural and less robotic. Users can interrupt, ask follow-up questions immediately, or speak without waiting for the AI to finish, leading to a smoother interaction.

First seen on Subvolts May 20, 2026. Updated Oct 18, 2026. Written by an AI model from the items above and reviewed against them weekly; report an error.