Event ·Multi-source

OpenAI 复盘：6 个月构建 GPT-Live 实时语音系统

OpenAI 复盘了 6 个月内构建 GPT-Live 语音系统的过程。GPT-Live 实现全双工对话，端到端延迟低于 1 秒。系统采用异步委托机制，快路径处理语音，慢路径调用 GPT-5.5 处理复杂逻辑。媒体前端从 Python 迁移到 Go 语言，降低了音频数据包延迟波动。自定义协议 WARP 基于 WebRTC 优化，将连接建立从 6 次网络往返缩减到 1 次。

Current read

OpenAI 自己拆解了 GPT-Live 怎么做到 1 秒内全双工对话，还开源了 WARP 协议，想搞实时语音的可以看这套架构。

6 sources82° AI heatLast updated 2026/8/4 09:52:44

Evidence

6 sources

Primary source小互

2026/8/4 09:52:44View source

Related source 1OpenAI Blog

2026/8/3 07:00:00View source

Related source 2Greg Brockman

2026/8/3 22:25:53View source

Related source 3ChatGPT

2026/8/4 18:31:35View source

Related source 4向阳乔木

2026/8/6 08:14:40View source

Related source 5Simon Willison’s Weblog

2026/8/4 23:58:24View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.

Open the story