--- title: How we built a realtime system for responsive voice AI... description: OpenAI's engineering team details how they built GPT-Live, a third-generation voice AI system, in six months. The system replaces the traditional turn-detector... canonical: https://daily.dev/posts/how-we-built-a-realtime-system-for-responsive-voice-ai-in-six-months-ea2qvvzg2 twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: How we built a realtime system for responsive voice AI in six months | daily.dev og:description: OpenAI's engineering team details how they built GPT-Live, a third-generation voice AI system, in six months. The system replaces the traditional turn-detector... og:url: https://daily.dev/posts/how-we-built-a-realtime-system-for-responsive-voice-ai-in-six-months-ea2qvvzg2 og:image: https://api.daily.dev/og/posts/EA2QVvZg2.png og:image:alt: How we built a realtime system for responsive voice AI in six months og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # How we built a realtime system for responsive voice AI in six months **[OpenAI](https://daily.dev/sources/openai)** · 14 min read · 0 upvotes · 0 comments ## Summary OpenAI's engineering team details how they built GPT-Live, a third-generation voice AI system, in six months. The system replaces the traditional turn-detector architecture with a full-duplex voice model that can listen and speak simultaneously. Key engineering decisions include: separating the media fast path from application logic, rewriting the media frontend in Go (replacing Python asyncio) for smoother frame delivery, using WebRTC as the transport layer, and developing WARP (WebRTC Abridged Roundtrip Protocol) to reduce session startup from six network round trips to one. The post also covers stateful inference with seamless model instance handoffs, dynamic context compaction without media interruption, asynchronous delegation to frontier models like GPT-5.5, and an 'Instant Connect' mechanism for near-instant session establishment. Production validation was done via shadow testing on real ChatGPT Voice traffic, which revealed capacity, geography, and observability challenges not visible in synthetic load tests. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What is GPT-Live and how does it differ from previous turn-based voice AI systems? GPT-Live is OpenAI's third-generation voice system that removes the turn detector from the audio path by using a full-duplex voice model that can listen and speak simultaneously. Previous systems relied on a separate turn detector model to decide when to respond, adding latency and risking cutting users off or feeling sluggish. GPT-Live consults frontier models like GPT-5.5 asynchronously for deeper reasoning without interrupting conversation flow. _Teams building conversational AI can track architectural shifts like this full-duplex voice model on daily.dev._ ### What is the WARP protocol and how much does it reduce WebRTC connection setup time? WARP (WebRTC Abridged Roundtrip Protocol) is a set of backward-compatible protocol improvements developed by OpenAI that reduces WebRTC media and data startup from six network round trips down to just one. It works by piggybacking the DTLS handshake over ICE (SPED), using the faster DTLS 1.3 handshake, pre-negotiating SCTP (SNAP), and pre-negotiating data channels instead of using DCEP. WARP support has already been added to libwebrtc and Pion. _Developers optimizing realtime media transport can follow protocol proposals like WARP through daily.dev._ ### Why did OpenAI rewrite its voice media frontend from Python asyncio to Go? OpenAI rewrote the media frontend and inference logic in Go to significantly improve the smoothness of audio frame delivery, replacing a previous Python asyncio implementation. The rewrite was substantial enough that the new Go system's p95 latency matched the old Python system's p50 latency, a meaningful gain for the sub-second responsiveness needed in live voice conversation. _Engineers weighing Go against Python for latency-sensitive services can compare real tradeoffs like this on daily.dev._ ## Similar posts on daily.dev - [OpenAI Details GPT-Live’s Architecture for Continuous Stateful Voice Interaction](https://daily.dev/posts/openai-details-gpt-live-s-architecture-for-continuous-stateful-voice-interaction-um8nuvvop) · InfoQ · 1 upvotes · 0 comments - [OpenAI's GPT-Live: ChatGPT voice that listens and talks](https://daily.dev/posts/openai-s-gpt-live-chatgpt-voice-that-listens-and-talks-nwap9djkb) · The Next Web · 9 upvotes · 3 comments - [OpenAI launches new voice intelligence features in its API](https://daily.dev/posts/openai-launches-new-voice-intelligence-features-in-its-api-ymtm53lpv) · TechCrunch · 0 upvotes · 0 comments --- Tags: [#performance](https://daily.dev/tags/performance), [#golang](https://daily.dev/tags/golang), [#webrtc](https://daily.dev/tags/webrtc), [#voice-ai](https://daily.dev/tags/voice-ai), [#real-time-systems](https://daily.dev/tags/real-time-systems) [View this post on daily.dev](https://daily.dev/posts/how-we-built-a-realtime-system-for-responsive-voice-ai-in-six-months-ea2qvvzg2) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"How we built a realtime system for responsive voice AI in six months","url":"https://daily.dev/posts/how-we-built-a-realtime-system-for-responsive-voice-ai-in-six-months-ea2qvvzg2","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/how-we-built-a-realtime-system-for-responsive-voice-ai-in-six-months-ea2qvvzg2"},"datePublished":"2026-08-03T20:25:06.904Z","dateModified":"2026-09-14T08:22:38.862Z","description":"OpenAI's engineering team details how they built GPT-Live, a third-generation voice AI system, in six months. The system replaces the traditional turn-detector...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/2deb97b48d21f51e90fdcac82f18e81a?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/2deb97b48d21f51e90fdcac82f18e81a?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"OpenAI","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"OpenAI","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/984b3632316f4afaa3d5213d5c64fe65","url":"https://daily.dev/sources/openai"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/how-we-built-a-realtime-system-for-responsive-voice-ai-in-six-months-ea2qvvzg2","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"performance,golang,webrtc,voice-ai,real-time-systems","timeRequired":"PT14M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"OpenAI","item":"https://daily.dev/sources/openai"},{"@type":"ListItem","position":3,"name":"How we built a realtime system for responsive voice AI in six months"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/how-we-built-a-realtime-system-for-responsive-voice-ai-in-six-months-ea2qvvzg2#faq","mainEntity":[{"@type":"Question","name":"What is GPT-Live and how does it differ from previous turn-based voice AI systems?","acceptedAnswer":{"@type":"Answer","text":"GPT-Live is OpenAI's third-generation voice system that removes the turn detector from the audio path by using a full-duplex voice model that can listen and speak simultaneously. Previous systems relied on a separate turn detector model to decide when to respond, adding latency and risking cutting users off or feeling sluggish. GPT-Live consults frontier models like GPT-5.5 asynchronously for deeper reasoning without interrupting conversation flow. Teams building conversational AI can track architectural shifts like this full-duplex voice model on daily.dev."}},{"@type":"Question","name":"What is the WARP protocol and how much does it reduce WebRTC connection setup time?","acceptedAnswer":{"@type":"Answer","text":"WARP (WebRTC Abridged Roundtrip Protocol) is a set of backward-compatible protocol improvements developed by OpenAI that reduces WebRTC media and data startup from six network round trips down to just one. It works by piggybacking the DTLS handshake over ICE (SPED), using the faster DTLS 1.3 handshake, pre-negotiating SCTP (SNAP), and pre-negotiating data channels instead of using DCEP. WARP support has already been added to libwebrtc and Pion. Developers optimizing realtime media transport can follow protocol proposals like WARP through daily.dev."}},{"@type":"Question","name":"Why did OpenAI rewrite its voice media frontend from Python asyncio to Go?","acceptedAnswer":{"@type":"Answer","text":"OpenAI rewrote the media frontend and inference logic in Go to significantly improve the smoothness of audio frame delivery, replacing a previous Python asyncio implementation. The rewrite was substantial enough that the new Go system's p95 latency matched the old Python system's p50 latency, a meaningful gain for the sub-second responsiveness needed in live voice conversation. Engineers weighing Go against Python for latency-sensitive services can compare real tradeoffs like this on daily.dev."}}]} ```