AI News

ChatGPT Is Retiring Advanced Voice: What Live Voice Changes for You

OpenAI is retiring Advanced and Standard Voice in ChatGPT on 11 December 2026. Here's what Live Voice changes, what stays and what to do before the switch.

NEXFRAME AI 12 min read
Nexframe AI featured image reading 'ChatGPT Is Retiring Advanced Voice' with an AI News label

OpenAI is removing two of the three voice modes in ChatGPT. On 11 December 2026, Advanced Voice and Standard Voice will be retired, and anyone still using them will be moved automatically to Live Voice, the ChatGPT voice experience powered by OpenAI's GPT-Live models. The one exception for now: live video and screen sharing will keep running on the older Advanced Voice until Live Voice can handle them.

This article explains what is changing, what Live Voice actually is, which plans get which version, what is confirmed and what is still unclear, and what you should do before the switch, with a few notes for people using ChatGPT on Nigerian mobile networks.

What Happened

OpenAI announced the change on 10 November 2026, in a post from its Newsroom account and a new help centre article titled "Retiring Advanced and Standard Voice in ChatGPT". Several outlets, including ANI (syndicated by The Tribune) and Interesting Engineering, reported the announcement that day and the next.

The key points, as set out by OpenAI and repeated in those reports:

  • Retirement date: Advanced Voice and Standard Voice will be removed on 11 December 2026. Until then, both remain available.
  • Automatic switch: If you have either older mode selected, ChatGPT will move you to Live Voice. The old options will disappear from Settings, then Voice.
  • Old conversations stay: Chats you started in the older modes remain in your history. You can reopen them and carry on by typing or speaking, but spoken replies will come from GPT-Live.
  • Video and screen sharing stay on Advanced Voice for now: OpenAI says these features will continue to run on Advanced Voice for eligible users until GPT-Live supports them. It has not given a date.
  • Custom GPTs: Voice chats with custom GPTs will keep using Advanced Voice until 11 December, which reports say is also the scheduled retirement date for custom GPTs for most users. Some Enterprise and government (FedRAMP) workspaces have a later date.

OpenAI's stated reason is simplicity. "This change will allow us to simplify the voice experience and focus on improving GPT-Live," the company said, according to ANI's report. OpenAI also said that voice usage in ChatGPT has nearly doubled since GPT-Live arrived, and that virtually all people using voice are already on it.

Why It Matters

Three voice modes become one

Until now, ChatGPT has quietly run three different voice systems, and many people did not know which one they were using.

  • Standard Voice chained three steps: one model turned your speech into text, a text model wrote the answer, and another model read it aloud. It was slower, but it used the same text model as your typed chats, which is why some long-time users preferred it for detailed answers.
  • Advanced Voice processed and produced audio directly, so it sounded more natural, but it waited for you to finish speaking before replying.
  • Live Voice (GPT-Live) is what OpenAI calls "full-duplex": it listens and speaks at the same time. It can give short acknowledgements while you talk, stay quiet while you think, and let you interrupt mid-sentence.

After 11 December, everyone gets the third option, apart from the video and screen-sharing exception.

The end of a "text model reads aloud" option

The biggest loss is for people who liked Standard Voice because its answers matched the depth of typed ChatGPT. Reports note that some Plus and Pro subscribers asked OpenAI to keep it. OpenAI says it will use feedback from users of the older modes to improve GPT-Live's response depth, continuity across conversations and personalisation, but it has not said how much those will improve or when.

Voice is becoming a front door to agents

GPT-Live is designed to hand heavier work to other models in the background. You can ask a voice question that needs a web search or a long piece of reasoning, and the conversation continues while that work runs. That makes voice less of a novelty and more of a way to control ChatGPT hands-free, which is why OpenAI wants everyone on one system.

What Live Voice Actually Is

Live Voice is not brand new. OpenAI introduced GPT-Live on 8 July 2026 and began rolling it out to ChatGPT users worldwide the same day.

How it launched

According to MacRumors' report on the launch, at that point:

  • Free users got GPT-Live-1 mini as their default voice model.
  • Go, Plus and Pro users got the full GPT-Live-1.
  • Video and screen sharing were not supported, and OpenAI said it was working on them.
  • It offered remastered voices, live translation and on-screen visual cards for things like weather, stocks and sports.

The October update

On 15 October 2026, OpenAI published a system card for GPT-Live-1.1, its updated voice model. The system card states that GPT-Live-1.1 is the default voice model for Plus and higher tiers, and that a smaller model, GPT-Live-1.1 Quip, is the default for other users. MIXED, citing OpenAI's Voice help page, reports that this means Free and Go users get Quip, and that Plus users get up to three hours a day on the full model, while Pro has no fixed cap.

The 1.1 update also added new voices for French, German, Hindi, Japanese, Korean and Mexican Spanish, plus a new English voice, with what OpenAI describes as more natural accents across languages.

What we could not confirm

OpenAI's Newsroom announced the 11 December date directly. The plan details above rely on OpenAI's GPT-Live-1.1 system card plus named reports, and a few points are still open:

  • We could not confirm exact daily voice limits for Free and Go users.
  • We found no OpenAI statement on whether any plan can keep an older mode after 11 December. Reports suggest no.
  • OpenAI has not given a date for video and screen sharing on GPT-Live.

Free, Go and Plus: Who Gets What

Plan Default voice model (per OpenAI's GPT-Live-1.1 system card and reports) What changes on 11 December
Free GPT-Live-1.1 Quip (smaller model) Older modes removed; you stay on Live Voice
Go GPT-Live-1.1 Quip (smaller model) Older modes removed; you stay on Live Voice
Plus GPT-Live-1.1 (full model), reported cap of three hours a day Older modes removed; video and screen sharing stay on Advanced Voice for now
Pro GPT-Live-1.1 (full model), reported no fixed cap Same as Plus

One detail worth noting: at the July launch, reports said Go users had the full GPT-Live-1. Under the October update, the system card's wording ("Plus and higher" for the full model) puts Go on Quip. If you chose Go partly for voice, test it now so you know what you are getting. Our ChatGPT Go vs Plus comparison covers the wider differences between the two plans.

What It Means for You

If you mostly type and rarely use voice

Very little changes. If you tap the voice button now and then, you are almost certainly already on Live Voice, since OpenAI says virtually all voice users are.

If you chose Standard Voice on purpose

This is the group most affected. After 11 December, your spoken questions will be answered by GPT-Live rather than your chosen text model reading its reply aloud. For long, detailed answers, such as a full study explanation or a business plan breakdown, you may get better results by typing and then using the read-aloud option on the response, if your app offers it, rather than relying on live voice.

If you use video or screen sharing

Keep using them as normal. OpenAI says these stay on Advanced Voice until GPT-Live supports them, so this is the one place the old mode lives on.

If you are a student or job seeker

Live Voice is well suited to speaking practice, because you can interrupt and it can stay quiet while you think. A Lagos graduate rehearsing for an interview could ask it to play a hiring manager and push back on weak answers. Once the real interview is done, our thank you email after an interview workflow picks up from there.

If you are in Nigeria

Three practical points.

Data usage. OpenAI does not publish how much mobile data a minute of Live Voice uses, and we have not tested it ourselves, so we will not guess a figure. Voice is a continuous audio stream, so it will generally use more data than typing the same question. A simple test: note your data balance, run a five-minute voice session, then check again. That tells you more about your own network and phone than any average would.

Nigerian languages. We found no OpenAI documentation listing Yoruba, Igbo, Hausa or Nigerian Pidgin as supported voice languages, and the October update's named language improvements did not include them. That does not mean they never work; it means quality is unconfirmed. Try a short conversation and judge for yourself. Our guide to using AI in Yoruba, Igbo, Hausa and Pidgin explains where AI tools tend to do well and where they struggle.

Paying for a plan. When ChatGPT Go launched in Nigeria in October 2025, TechCabal reported a price of ₦7,000 a month. Prices can change, so check the "Upgrade plan" screen in your own app before paying. Remember the table above: on the system card's wording, Go and Free share the smaller Quip voice model, while Plus gets the full one.

A worked example

Chiamaka (an invented example) runs a small catering business in Enugu and uses Standard Voice while cooking to draft supplier messages and price quotes. After 11 December, she will be on Live Voice. Before the switch, she spends ten minutes testing it with her usual tasks: "Draft a WhatsApp message to my rice supplier asking for a new price on ten bags." If the replies feel too short, she can say so in the conversation, or switch to typing for longer drafts. If she finds that voice uses more data than she expected on her mobile plan, she saves longer sessions for when she is on Wi-Fi.

What to Do Before 11 December

You do not have to do anything; the switch is automatic. But a few minutes now can save frustration later.

  1. Check which mode you are on. Go to Settings, then Voice. If you see Advanced or Standard selected, you will be moved.
  2. Try Live Voice with your real tasks. Use the same kinds of questions you normally ask. Note anything that works worse, such as answer length or accuracy on a topic you know well.
  3. Pick a voice you like. The 1.1 update added new voices. Reports say you can choose one under Settings, then Voice, though options vary by region and account.
  4. Save anything important from old voice chats. Your history stays, according to OpenAI, but if a voice conversation contains something you rely on, such as a recipe, a plan or a set of instructions, copy it into a note now. It is a habit worth having anyway.
  5. Test data usage on your network. Run the five-minute test described above so you know what a typical session costs you.
  6. Learn the interruption habit. Live Voice is built to be interrupted. If an answer goes in the wrong direction, just start talking. This is one of the ChatGPT features most people never use, and it makes voice much faster.
  7. Leave video and screen sharing alone. They keep working on Advanced Voice for now.

A Word on Voice Scams

As AI voices sound more human, voice-based fraud gets easier to pull off. None of this is specific to ChatGPT. But the wider trend matters: criminals use cloned voices to pose as relatives, bosses or bank staff and ask for urgent transfers.

Agree a family code word, hang up and call back on a number you already know, and never send money because a voice sounds familiar. Our guide to AI voice cloning scams goes through the warning signs in detail.

What to Watch Next

  • Video and screen sharing on GPT-Live. This is the last reason Advanced Voice exists. When OpenAI adds them to Live Voice, expect Advanced Voice to disappear completely.
  • Answer depth. OpenAI has promised improvements in response depth, continuity and personalisation. Whether former Standard Voice users are satisfied will depend on how much that improves.
  • Plan limits. Watch for any change to the reported Plus cap, or to which model Go users get.
  • Competitors. Voice is now a main battleground among AI assistants. Anthropic and Google both had significant model news this week; see our coverage of Claude Fable 5.2 and Gemini 4 Hydrogen. For a side-by-side look at how the major assistants handle voice, read our AI voice assistants comparison.

Final Thoughts

This is a tidy-up rather than a revolution. Most people already use Live Voice without knowing its name, and on 11 December the older options simply disappear. The people who should pay attention are those who deliberately chose Standard Voice for its fuller answers, anyone who relies on video or screen sharing, and anyone watching their data or their plan budget closely. Spend ten minutes testing Live Voice with your real tasks before the switch, and you will know exactly what you are getting.

Frequently Asked Questions

When is ChatGPT retiring Advanced Voice?

OpenAI says Advanced Voice and Standard Voice will be removed from ChatGPT on 11 December 2026. The company announced the change on 10 November 2026.

What is Live Voice in ChatGPT?

Live Voice is ChatGPT's voice mode powered by OpenAI's GPT-Live models, introduced in July 2026. It listens and speaks at the same time, so you can interrupt it, and it can pass harder tasks to other models in the background while you keep talking.

Will I lose my old voice conversations?

No. OpenAI says conversations started in the older modes stay in your chat history, and you can continue them by typing or speaking. Spoken replies will come from GPT-Live.

Can I still use video and screen sharing in ChatGPT voice?

Yes, for now. OpenAI says live video and screen sharing will keep running on Advanced Voice until GPT-Live supports them. It has not given a date.

Do free ChatGPT users get Live Voice?

Yes. According to OpenAI's GPT-Live-1.1 system card, users below the Plus tier get a smaller model called GPT-Live-1.1 Quip by default, while Plus and higher get the full GPT-Live-1.1.

Does ChatGPT voice understand Yoruba, Igbo, Hausa or Pidgin?

OpenAI has not documented support for these languages in voice, as far as we could find. Results may vary, so test a short conversation in your language before relying on it.

Share

Comments (0)

Sign in to post a comment.

  • Be the first to comment.