Introduction

Two of ChatGPT's voices have a month left. On November 10, 2026, OpenAI said it will retire Advanced voice and Standard voice from ChatGPT Voice on December 11, 2026. After that, every voice chat runs on GPT-Live, the full-duplex model it launched in July.

This is a change to a closed consumer product, not a new model. It shows up in the ChatGPT release notes, a dedicated Help Center article, and an OpenAI Newsroom post on X. If you use voice on chatgpt.com or in the iOS and Android apps, you'll be moved to Live automatically.

What's Changing

  • Retiring: Advanced voice and Standard voice in ChatGPT Voice.
  • Date: December 11, 2026. Until then, you can still pick either one in Settings > Voice.
  • What happens on the day: both options disappear from the voice picker, and anyone who still has one selected is switched to Live.
  • Replacement: Live, powered by GPT-Live-1.1, or GPT-Live-1.1 Quip on lower tiers.
  • The exception: live video and screen sharing keep running on Advanced voice until GPT-Live supports them. Live can't do either today.
  • Custom GPTs: voice in custom GPTs stays on Advanced until December 11. That's also when custom GPTs themselves are retired.
  • Your history: old Advanced and Standard voice chats stay in your history, and you can continue them in text or in Live.
  • Why: OpenAI says voice usage has nearly doubled since GPT-Live launched, and "virtually all" ChatGPT voice users already use it.

Three Voice Stacks, Explained

OpenAI's FAQ describes the three modes very clearly. It's also a nice short history of how voice AI has changed.

  • Standard voice chained three models together. A speech-to-text model transcribed you, a text model wrote a reply, and a text-to-speech model read it back. It was slow, and tone and emotion were lost between steps. Its audio was deleted after transcription.
  • Advanced voice was a native speech-to-speech model. It heard and spoke audio directly, but it was turn-based, so it waited for you to stop talking before it answered.
  • GPT-Live is full-duplex, which means it listens and talks at the same time. It can say "mhmm" while you talk, handle quick back-and-forth, stay quiet while you think, and hand harder questions or agent tasks to OpenAI's text models while the conversation keeps going.

What Live Gives You Today

  • Plans and limits (rolling 24 hours): Free gets limited GPT-Live-1.1 Quip. Go gets 3 hours of Quip. Plus gets 3 hours of GPT-Live-1.1. Pro $100 gets 15 hours, and Pro $200 is unlimited, subject to abuse guardrails.
  • Business and Enterprise: Business Standard includes 3 hours and Business Premium 15 hours, then extra use costs 1.25 credits per minute. Enterprise on USD usage pricing pays $0.05 per minute.
  • Features: web search, memory, plugins and connected apps, images and text in the same chat, and Voice in Work.
  • Voices: nine core voices, from Arbor to Vale, plus language-specific Live voices for Brazilian Portuguese, French, German, Hindi, Japanese, Korean and Mexican Spanish, and Bubbie in English.
  • Missing pieces: no video, no screen sharing, no preset personalities, and no exact playback-speed control. You can still ask it to talk faster or slower.

Pro Tip

If you rely on Advanced for something Live doesn't do yet, like pointing your camera at a whiteboard, keep doing that in the mobile app. That path stays on Advanced after December 11. For everything else, switch to Live now in Settings > Voice, while you can still switch back and compare. On a busy desk, tell it at the start: "Wait until I ask you to respond." That stops a full-duplex model from jumping in while you think out loud.

The Reaction

Many replies to the Newsroom post ask OpenAI to keep Standard voice, or to open-source its text-to-speech voices if it's too costly to run. OpenAI's help article points people to its guidance on navigating model transitions, but it says nothing about releasing the old voices. Some people really do prefer a calm, turn-by-turn voice that never talks over them, and Live is built to behave the other way.

Where It Fits

This matches the rest of the industry. Real-time voice is moving from cascaded and turn-based systems to full-duplex models. We've covered rivals like Fish Audio Live 3, and OpenAI's own GPT-Live 1.1 voices and prompt-to-voice API. The notice only covers ChatGPT Voice. It doesn't mention any change to OpenAI's API.

Conclusion

The voice that waits for you to finish is going away, and one that can listen while it talks is taking over. For most people that's an upgrade. If you liked Standard's slower pace, though, now is the time to try Live and tell OpenAI what's missing.

—Anabel ♡