ChatGPT Advanced Voice Retired: What Changes on December 11, 2026

ChatGPT Advanced Voice Retired: What Changes on December 11, 2026

On 10 November 2026, OpenAI gave millions of ChatGPT users a date to circle: ChatGPT Advanced Voice retired on 11 December 2026, along with the older Standard Voice mode. From that Friday, the two older voice options disappear from Settings > Voice, and everyone who talks to ChatGPT out loud will be moved to Live voice, powered by the company’s newest voice model, GPT-Live.

If you use voice to brainstorm, practise a language, dictate ideas on a walk or talk through a problem hands-free, this change reaches you directly — and the smart move is to understand it now, while you still have a month to get comfortable with the replacement. This guide explains exactly what is retiring, what GPT-Live does differently, what stays the same, how much voice time each plan gets, and the simple steps to take before 11 December so the switch costs you nothing.

Table of Contents

ChatGPT Advanced Voice retired — a woman talking to her phone using a voice assistant
ChatGPT Advanced Voice retired on 11 December 2026 — from that day, every voice conversation in ChatGPT runs on GPT-Live instead.

What Is Happening on December 11?

The short version is simple. ChatGPT Advanced Voice retired is not a rumour or a leaked plan: it is OpenAI’s own announcement, published in the ChatGPT release notes and in a dedicated Help Center article on 10 November 2026. On 11 December 2026, OpenAI will remove Advanced Voice and Standard Voice from ChatGPT Voice.

Anyone who had picked one of the two older modes will be switched to Live voice automatically. You do not need to do anything for the switch to happen — but doing nothing also means your first conversation on the new system will be a surprise, and for people who rely on voice every day, a surprise is the worst way to meet a new interface.

To understand why this matters, it helps to know what the three modes actually are, because they are not three skins on the same engine. Standard Voice was ChatGPT’s original spoken mode: it chained three separate models together — one to turn your speech into text, a large language model to write the reply, and a third to read that reply aloud.

It worked, but it was slow, and information was lost every time the conversation jumped between voice and text. ChatGPT Advanced Voice, launched later, processed audio natively in a single model, which made it faster and more natural — but it was still turn-based, meaning it waited for you to finish speaking before it answered. GPT-Live, introduced in July 2026, is the third generation: a full-duplex voice model that listens and speaks at the same time, the way a person in a real conversation does.

OpenAI’s stated reason for the change is that the new model has already won. The company says voice usage has nearly doubled since GPT-Live arrived in July, and that virtually everyone who uses voice in ChatGPT now uses GPT-Live. With the audience already moved, keeping two older voice engines running — each with its own infrastructure, maintenance and improvement work — no longer makes sense. The retirement lets OpenAI put all of its voice effort into one model instead of splitting it across three.

The notice period is 31 days, from 10 November to 11 December. That is short for a change that affects how millions of people talk to the most popular chatbot in the world, and it is one reason the announcement drew a fast reaction. But the date is also the whole point of this guide: you have a month, and a month is plenty if you use it to try the new mode on your own terms instead of being switched on OpenAI’s schedule.

ChatGPT Advanced Voice Retired: What Exactly Is Retiring — and What Is Not

Headlines about a retirement are easy to misread, so here is the precise picture. The ChatGPT Advanced Voice retired announcement covers the two selectable voice conversation modes — the ones you start with the waveform icon and choose in Settings > Voice. Everything else around voice in ChatGPT is either staying or changing on a different schedule.

FeatureWhat happens on 11 December 2026
Standard VoiceRetired. Can no longer be selected in Settings > Voice.
Advanced VoiceRetired as a selectable mode. Still powers live video and screen sharing for now.
Live voice (GPT-Live)Becomes the default and only selectable voice mode.
Past voice conversationsStay in your chat history. Can be continued in text or with GPT-Live.
Read AloudUnchanged. The button that reads written replies aloud is not part of this retirement.
Live video and screen sharingKeep running on Advanced Voice until GPT-Live gains the capability. No end date announced.
Voice inside custom GPTsKeeps using Advanced Voice until custom GPTs themselves are retired on 11 December 2026.

The most important row in that table is the one people miss: voice itself is not going away. Nothing in OpenAI’s announcement suggests ChatGPT is dropping spoken conversation. The opposite is true — the company is consolidating voice onto its newest model because voice usage has grown so fast. If you only ever tap the waveform icon and talk, your routine survives; what changes is the engine answering you, and the way the conversation feels.

The second row worth pausing on is live video and screen sharing. If you point your camera at a whiteboard, a broken appliance or a document and talk ChatGPT through what you see, that feature still runs on Advanced Voice after 11 December, because GPT-Live cannot do it yet.

OpenAI has given no date for when GPT-Live will take over, which means ChatGPT Advanced Voice is not fully switched off on 11 December — it is retired as a choice, while quietly continuing behind one feature. That is worth knowing if you are one of the people who uses video mode, because your experience may not change at all on the day, and then change later without a second announcement.

Custom GPTs are the other exception with a deadline of their own. Voice chats inside a custom GPT keep using Advanced Voice until 11 December, when custom GPTs themselves are scheduled for retirement. If you built or rely on a custom GPT that you talk to, the voice change and the custom GPT retirement land on the same day — two changes, one date, and a reason to check whether the GPT you use has a future beyond December.

What Is GPT-Live?

GPT-Live is OpenAI’s latest voice model, and it is the reason the older modes are being retired. It was introduced in July 2026, and the company describes it as the model that “combines natural conversation with frontier intelligence”. The technical difference is full-duplex operation: GPT-Live can listen and speak at the same time. The human difference is that a conversation with it feels less like leaving voicemails and more like talking to someone in the room.

How Full-Duplex Changes the Conversation

With Standard and Advanced Voice, the rhythm was always the same: you speak, you stop, the model thinks, the model speaks, and then it waits for you again. That turn-taking is fine for a quick question, but it breaks down in the situations where voice is most useful — thinking aloud, changing your mind halfway through a sentence, or answering a question with a question. GPT-Live removes the wait. It can hear you while it is talking, so you can interrupt it mid-sentence, correct yourself, or say “wait, actually” and have it respond to the correction instead of ploughing on with the old thread.

It also behaves more like a listener. OpenAI’s Help Center says GPT-Live can show it is paying attention with short phrases like “mhmm” or “yeah”, engage in quick back-and-forth, or simply stay quiet while you need a moment to think. That sounds small, but it is the difference between a voice tool that demands your full attention and one you can think alongside. If you use voice to work through an idea — dictating a draft, talking through a decision, rehearsing a difficult conversation — the model staying quiet while you think is a genuine improvement, not a gimmick.

There is a second, quieter upgrade. GPT-Live can delegate complex questions or agentic tasks to OpenAI’s frontier text models while keeping the conversation going. In practice, that means you can ask it to do something involved — look something up, work through a multi-step problem, start a task in the background — and keep talking while it works, instead of the conversation freezing while a single model grinds through the job.

OpenAI has said it is still improving GPT-Live in exactly the areas long-time Advanced Voice users care about: the depth and detail of responses, continuity across conversations, and better personalisation. Those are honest admissions that the new model is not yet better at everything, and they are worth keeping in mind if your first GPT-Live conversations feel thinner than the ones you are used to.

GPT-Live-1.1: New Voices for Life After ChatGPT Advanced Voice Retired

The model you will meet on 11 December is not quite the one that launched in July. GPT-Live-1.1 has been rolling out since 15 October 2026, and it adds 13 new voices to the selection, along with noticeably better accents in French, German, Hindi, Mexican Spanish, Korean and Japanese. For the large number of ChatGPT users who talk to the app in a language other than English — or in English with an accent the older models handled clumsily — that is the most practical improvement in the whole announcement.

The voice selection matters more than it might seem, because voice is personal in a way text is not. People who use Advanced Voice every day have a favourite voice, a speaking pace they are used to, and a rhythm they have tuned to the model’s pauses. GPT-Live-1.1’s wider selection is OpenAI’s answer to that: if the default voice grates, there are more options to try, and the new accents mean the model sounds less like a generic American voice reading your words back to you. Spend a few minutes in Settings > Voice before December trying the new voices — it is the single easiest way to make the forced switch feel like a choice.

ChatGPT Advanced Voice retired — a man talking to a smart speaker at home
Talking to an AI is now an everyday habit at home — and the ChatGPT Advanced Voice retired change affects everyone who uses voice in the ChatGPT app.

The Timeline: What Happens and When

OpenAI published the retirement as a three-stage schedule, and it is worth reading as a schedule rather than a single event, because the middle stage is the useful one.

WhenWhat happens
10 November 2026Users are notified about the retirement, in the release notes and in the app.
10 November – 10 December 2026Transition period. You can keep using Advanced or Standard Voice if you have picked one. This is the window to try GPT-Live on your own terms.
11 December 2026Advanced and Standard Voice can no longer be selected in Settings > Voice. Anyone still on them is moved to Live voice automatically. Custom GPTs are also retired on this date.

One detail OpenAI did not specify is the time of day or time zone for the cut-off. 11 December 2026 is a Friday, and the announcement gives a date, not an hour. In practice that means you should treat the evening of 10 December as your real deadline: if you have a voice routine you depend on — a morning dictation habit, a language-practice session, a regular call with a custom GPT — assume it switches when you wake up on the 11th, and be ready for it.

The transition period is also the only part of this story you control. Between now and 11 December, both the old modes and the new one are available, which makes this the one month in which you can compare them side by side on the same questions, in the same situations, with the same voice habits.

That comparison is worth doing deliberately: ask GPT-Live the questions you normally ask Advanced Voice, interrupt it once, let a silence hang, and see how it handles each. If it handles them well, the switch is a non-event. If it does not, you will know exactly what to complain about — or what workaround to set up — before the old mode disappears.

What Stays the Same After the Switch

A retirement announcement invites worst-case thinking, so it is worth listing what does not change on 11 December, because the list is longer than the list of things that do.

  • Your conversation history stays. Every voice conversation you have had remains in your ChatGPT history, and you can continue any of them in text or with GPT-Live. Nothing is deleted, and no export is needed to keep your chats — although exporting is still a good habit, and OpenAI’s Help Center links to its data-export page for anyone who wants a copy.
  • Read Aloud is untouched. The button that reads a written reply out loud is a separate feature from the voice conversation modes, and OpenAI’s Help Center confirms it is not part of this retirement. If you use Read Aloud to listen to long answers while you drive or cook, it keeps working exactly as before.
  • Voice itself is still there. The waveform icon, the hands-free mode, the ability to talk instead of type — all of it survives. Only the engine behind it changes.
  • Your plan does not change. The retirement is not a price change or a plan reshuffle. Your subscription, your limits and your billing stay as they are — although the voice allowances inside each plan are worth checking, because they differ more than most people realise, as the next section shows.
  • Live video and screen sharing keep working. They will keep running on ChatGPT Advanced Voice until GPT-Live can handle them, with no end date announced. If video mode is your main use, your experience may not change on 11 December at all.

The honest summary is that this is a change of engine, not a removal of a feature. The risk is not that voice disappears; it is that the new engine feels different in the specific ways you use it most — the depth of its answers, how well it remembers what you said five minutes ago, how natural its interruptions feel. Those are real differences, and they are the reason to test GPT-Live now rather than discover them on a busy morning in December.

ChatGPT Voice Limits by Plan in 2026

One thing the retirement announcement does not change is how much voice time you get — but the switch to GPT-Live makes those limits matter more, because Live voice is the only mode left, and every plan’s allowance is now an allowance for the same model. OpenAI’s help documentation lists the following Live voice allowances, measured over a rolling 24-hour period, with a notification when you approach the limit.

PlanPrice (US)Live voice allowance
Free$0Limited access to GPT-Live-1.1
Go$8/month3 hours with a Live model
Plus$20/month3 hours with a Live model
Pro ($100)$100/month15 hours with a Live model
Pro ($200)$200/monthUnlimited, subject to abuse guardrails

A few things in that table deserve a closer look. First, the free tier’s “limited access” is genuinely limited: free users get a taste of GPT-Live-1.1, not a daily voice habit. If you talk to ChatGPT every day without paying, the retirement will feel sharper, because the mode you lose (Standard Voice) was the one with the most generous free access, and the mode you gain is metered.

Second, Go and Plus share the same three-hour allowance, which makes Go — at $8 a month — the cheapest way to get a meaningful daily voice allowance if voice is your main use and you do not need Plus’s other extras. Third, the jump from Plus to the $100 Pro plan is a five-fold increase in voice time, from 3 hours to 15, which is the plan to look at if you use voice for hours a day — long dictation sessions, language practice, or working through problems out loud. And the $200 Pro tier’s “unlimited” still carries abuse guardrails, so it is unlimited in the normal sense of the word, not in the absolute sense.

There is a practical point hidden in the rolling 24-hour measurement. Your allowance is not a calendar-day allowance that resets at midnight; it is a rolling window, so heavy use in the evening eats into the next morning. If you hit the limit, ChatGPT notifies you, and the only fix is to wait for the window to roll forward — or to switch to text. That is worth knowing before 11 December, because the users most likely to hit the wall are the ones who used Advanced Voice for long, unmetered sessions and never had to think about a clock.

For the wider context on what OpenAI is charging for and where ChatGPT fits into business use, our ChatGPT Enterprise guide covers the business tiers, and our Businesses in ChatGPT explainer covers how companies are now appearing inside the app itself.

GPT-Live vs Advanced Voice vs Standard Voice

The three modes are three generations of the same idea, and the comparison shows why OpenAI is confident enough to retire two of them at once — and also where the new model still has something to prove.

Standard VoiceAdvanced VoiceLive voice (GPT-Live)
How it worksThree chained models: speech-to-text, language model, text-to-speechOne turn-based voice model, processes audio nativelyFull-duplex model — listens and speaks at the same time
Conversation feelSlow, with pauses between turnsFaster, but rigid back-and-forthNatural back-and-forth, interruptions work
Can you interrupt?NoOnly awkwardly, between turnsYes, mid-sentence
Background listeningNoNoYes — can acknowledge while you talk, stay quiet while you think
Video and screen sharingNoYes — still the only mode that does thisNot yet
Multi-taskingNoNoYes — can delegate to frontier text models in the background
Status after 11 Dec 2026RetiredRetired as a choice; still runs video modeDefault and only selectable voice

Read as a table, the retirement looks inevitable: GPT-Live wins on every row except video, where it loses by default because it cannot do the job yet. Read as a user experience, the picture is more mixed.

ChatGPT Advanced Voice users who valued its depth — the long, detailed spoken answers, the way it held a thread across a long conversation — have reason to be cautious, because OpenAI itself lists depth, continuity and personalisation as areas it is still improving in GPT-Live. A model that can interrupt you is not automatically a model that answers as well. The table tells you what the new model can do; only using it tells you whether it does the things you care about as well as the old one did.

There is also a comparison worth making outside ChatGPT. Google’s Gemini Live is the closest rival, and it already works in a similar full-duplex way inside the Android app — if you are on Android and the GPT-Live switch disappoints you, Gemini Live is the most direct alternative, and it costs nothing to try.

Anthropic’s Claude has voice on mobile, desktop and web, though it is a smaller part of the Claude experience than voice is of ChatGPT’s. And for people whose main use is dictation or transcription rather than conversation, a dedicated tool will always beat a chatbot at that one job. The point is not that you must leave ChatGPT — it is that 11 December is a natural moment to check whether the tool you talk to every day is still the best one for the way you actually use it.

ChatGPT Advanced Voice retired — setting up a voice assistant speaker at home
Voice assistants are now part of the furniture — and the ChatGPT Advanced Voice retired switch is the biggest change to how ChatGPT talks since Advanced Voice launched.

How to Prepare Before December 11

You have a month, and the preparation is mostly a matter of trying the new mode while the old one is still there to fall back on. Here is a practical checklist, in the order that matters most.

  1. Switch to Live voice for one day this week. Open ChatGPT, tap the waveform icon, and if you are still on ChatGPT Advanced Voice or Standard Voice, change the mode in Settings > Voice to Live. Use it for your normal routine — not a test script, your real questions — and notice what feels different. One day of real use tells you more than any review.
  2. Try interrupting it. Start a long answer, then cut in mid-sentence with a correction or a new question. This is the feature the old modes could not do, and it is the one that changes the rhythm of a conversation most. If it works well for you, the switch is an upgrade. If it feels chaotic, you will want to slow down and speak in shorter turns until you adjust.
  3. Pick a voice you like. GPT-Live-1.1 added 13 new voices, and the accent improvements cover French, German, Hindi, Mexican Spanish, Korean and Japanese. Spend five minutes in Settings > Voice listening to the options. A voice you like makes a forced switch feel like a choice; a voice you dislike makes every conversation slightly worse for no reason.
  4. Check your plan’s voice allowance. Look at the table above and work out whether three hours a day covers you. If you are on the free tier and you talk to ChatGPT daily, expect to hit the limit — and decide now whether Go at $8 a month is worth it, rather than discovering the limit on a day you need voice most.
  5. Test the things you actually use voice for. Language practice, dictation, brainstorming, talking through a problem, asking questions while your hands are busy — each of these stresses a different part of the model. Test the one you care about most, not the one in the demo. If you dictate long passages, check whether GPT-Live keeps up without cutting you off. If you practise a language, check the accent handling in that language specifically.
  6. Export your important voice chats. Your history stays, and OpenAI says conversations can be continued in text or with GPT-Live. But “stays in history” is not the same as “easy to find”, and a chat you care about — a long brainstorm, a language-practice thread, a dictation you never wrote down — deserves an export. OpenAI’s Help Center links to its data-export page; use it for anything you would regret losing.
  7. Check your custom GPTs. If you talk to a custom GPT regularly, remember that custom GPTs themselves are retired on 11 December, on the same day as the voice modes. Work out now whether the GPT you use has a replacement, or whether its job needs to move into a regular ChatGPT conversation.
  8. If you use video or screen sharing, relax — for now. That feature keeps running on Advanced Voice after 11 December. But note the “for now”: OpenAI has given no date for GPT-Live taking it over, so keep an eye on the release notes. The second switch, when it comes, will not get a month’s notice in the same way, because there will be no mode left to retire — just a quiet engine change behind a feature you already use.

None of this is difficult, and none of it costs anything except a little attention. The people who will have a bad 11 December are the ones who never opened the new mode until the old one was gone. A month of overlap is a gift — use it.

Why OpenAI Is Retiring the Older Voice Modes

Companies retire features for two kinds of reasons: because the feature failed, or because it succeeded so well that it was replaced. This is the second kind, and the numbers OpenAI gives make the logic plain. Voice usage has nearly doubled since GPT-Live arrived in July 2026, and virtually everyone who uses voice in ChatGPT now uses GPT-Live. When a new product already has the audience, the old products become maintenance cost: three voice engines to run, three sets of bugs to fix, three improvement roadmaps to staff, for an audience that has mostly left two of them behind.

There is also a strategic reason that matters for the longer term. GPT-Live is not just a better voice mode; it is a different kind of system. Because it can delegate to frontier text models in the background while keeping a conversation going, it is the voice layer for the agentic ChatGPT that OpenAI is building — the one that can shop, book, research and act, not just answer.

Our Businesses in ChatGPT explainer covers one piece of that shift: the app is becoming a place where you do things, not just a place where you ask things. A turn-based voice model cannot carry that kind of product. A full-duplex one that works in the background can. Retiring the older modes is OpenAI clearing the decks for the version of ChatGPT it actually wants to build.

That is also why the retirement is worth reading alongside the other November changes at OpenAI. The company has been moving fast — new models, new features, new ways to use ChatGPT — and voice is the interface most users touch every day. Consolidating it onto one model is the kind of unglamorous infrastructure work that makes the flashier launches possible. It is not the most exciting announcement of the month, but it may be the one that affects the most people.

The Backlash — and Why the Date Might Still Move

Not everyone is happy. The announcement drew quick user backlash, reported by outlets including Neowin, from people who prefer ChatGPT Advanced Voice’s style and depth and do not want to be moved to a model they see as chattier and less substantial. The complaints are familiar to anyone who has watched a popular product change: the new version is faster but shallower, the old version understood me better, why fix what was not broken. Some of it is nostalgia for a tool people have built a daily habit around, and some of it is a real, specific complaint about response depth that OpenAI itself has acknowledged by listing depth as an area it is still improving.

There is also a precedent that makes the backlash worth taking seriously. OpenAI announced the retirement of Standard Voice once before, in September 2025 — and then backed off after user feedback, saving the mode. The company has changed course on voice before, and it knows the audience is watching for it to do so again. That history does not mean 11 December will move. This time OpenAI has printed a fixed date in its release notes, given a month’s notice, and tied the change to a model that — by its own numbers — already has nearly all voice users. The company has more reason to hold the date this time than it did last September.

Even so, the honest advice is to treat 11 December as the plan, not a certainty. If you are the kind of user who would rather keep Advanced Voice, the useful move is not to hope the date moves — it is to make your preference visible in the normal channels, and in the meantime to get comfortable with GPT-Live anyway, so that either outcome leaves you ready. And if you are a heavy voice user, keep an eye on the release notes as December approaches: if the date does shift, it will shift there first.

ChatGPT Advanced Voice retired — a person wearing headphones on a video call
Voice is now how many people work, learn and practise languages — the ChatGPT Advanced Voice retired change lands in the middle of those daily routines.

Alternatives If You Do Not Like GPT-Live

If you try GPT-Live during the transition month and decide it is not for you, you are not stuck. The voice-assistant market in late 2026 is crowded, and several of the alternatives are free to try.

  • Gemini Live (Google). The closest direct replacement: full-duplex voice conversation, available in the Gemini app on Android and iOS, with the advantage of deep integration into Google’s services — calendar, email, maps — if you live in that ecosystem. If your main use is asking questions and getting things done hands-free, Gemini Live is the first alternative to try.
  • Claude voice (Anthropic). Claude has voice on mobile, desktop and web, and it suits people who value careful, detailed spoken answers over speed. Anthropic has also been generous with usage this month — our Claude free usage reset guide explains the free refill every paying Claude subscriber received in November — so it is a cheap moment to try Claude’s voice mode properly.
  • ElevenLabs Conversational AI. If you want a voice agent you can shape — a specific voice, a specific personality, a specific job — ElevenLabs’ conversational platform is the most flexible option, and it is built for developers and businesses rather than casual chat. It is overkill for asking questions on a walk, and exactly right if you are building a voice product.
  • Dedicated dictation and transcription tools. If your real use is dictation — turning speech into text you can edit — a dedicated tool will beat any chatbot at that one job, because it is the whole job. ChatGPT’s voice modes are for conversation; if you never actually want a reply, stop paying for one.

For most people, though, the honest answer is simpler: try GPT-Live properly first. The users who dislike it most on day one are often the ones who never changed the voice, never tested an interruption, and compared it to a year of habit with the old mode. Give it a real week of your real routine before you decide it is worse. And if it is worse for the specific thing you do — the long answers, the continuity, the depth — say so. OpenAI has shown before that it listens when enough voice users speak up.

ChatGPT Advanced Voice retired — a driver using voice control in the car
Hands-free voice is now part of driving, cooking and working — the ChatGPT Advanced Voice retired switch changes the voice behind all of those moments.

Frequently Asked Questions

When is ChatGPT Advanced Voice retired?
11 December 2026. OpenAI announced the date on 10 November 2026, in the ChatGPT release notes and in a Help Center article. From that day, Advanced Voice and Standard Voice can no longer be selected in Settings > Voice, and users on either mode are moved to Live voice automatically.

Is ChatGPT voice going away completely?
No. Only the two older modes — Advanced Voice and Standard Voice — are retiring. Voice itself continues, powered by GPT-Live, which becomes the default and only selectable voice mode. The Read Aloud button on written replies is also unchanged.

What replaces Advanced Voice?
Live voice, powered by GPT-Live, OpenAI’s full-duplex voice model introduced in July 2026. It can listen and speak at the same time, so you can interrupt it, and it can acknowledge you while you talk or stay quiet while you think. GPT-Live-1.1, rolling out since 15 October 2026, adds 13 new voices and better accents in several languages.

What happens to my old voice conversations?
They stay in your ChatGPT history and can be continued in text or with GPT-Live. Nothing is deleted. If a conversation matters to you, you can also export your data through OpenAI’s data-export page.

Will live video and screen sharing still work after 11 December?
Yes. They will keep running on Advanced Voice until GPT-Live gains that capability. OpenAI has not given a date for the changeover, so video mode users may notice no change at all on 11 December.

How much Live voice time does my plan include?
Free users get limited access to GPT-Live-1.1. Go and Plus get 3 hours with a Live model per rolling 24 hours. The $100 Pro plan gets 15 hours, and the $200 Pro plan is unlimited subject to abuse guardrails. ChatGPT notifies you as you approach the limit.

Could OpenAI delay the retirement?
Possibly. OpenAI announced a Standard Voice retirement in September 2025 and then reversed it after user feedback. This time the company has printed a fixed date and says almost all voice users are already on GPT-Live, so a delay is less likely — but treat 11 December as the plan, not a certainty, and watch the release notes as the date approaches.

What about voice inside custom GPTs?
Voice with custom GPTs keeps using Advanced Voice until custom GPTs themselves are retired on 11 December 2026, the same day as the voice modes.

The Verdict

ChatGPT Advanced Voice retired is a date, not a disaster. On 11 December 2026, the two older voice modes leave Settings > Voice, and GPT-Live becomes the only way to talk to ChatGPT out loud. Voice itself is not going anywhere — usage has nearly doubled since GPT-Live arrived, and OpenAI is consolidating onto the model its users have already chosen.

The new model is genuinely better at the mechanics of conversation: it listens while it speaks, it lets you interrupt, it stays quiet while you think, and it can work on things in the background while you keep talking. It is still catching up on depth, continuity and personalisation, and it cannot yet do live video or screen sharing — which is why Advanced Voice quietly survives behind that one feature.

The right response is not to panic and not to ignore it either. Use the transition month. Spend one day on Live voice, interrupt it once, pick a voice you actually like, check whether your plan’s three hours cover you, and export the conversations that matter. If GPT-Live wins you over, 11 December will pass unnoticed. If it does not, you will know exactly why — and you will have a month’s head start on the alternative. Either way, the worst outcome is the one that costs nothing to avoid: waking up on a Friday in December to a voice assistant you have never tried, on a morning you needed it to just work.

Sources: OpenAI Help Center, “Retiring Advanced and Standard Voice in ChatGPT” (help.openai.com); OpenAI ChatGPT release notes, 10 November 2026. Plan allowances as listed in OpenAI’s help documentation. GPT-6.2 context: our GPT-6.2 explained guide.

Leave a Comment

Your email address will not be published. Required fields are marked *