Key Takeaways
- GPT-Live-1.1 is OpenAI's updated voice model for Voice in ChatGPT, Codex and dots, announced October 15, 2026 and rolling out to all users the same day.
- It adds new voices and improved accents in French, German, Hindi, Mexican Spanish, Korean and Japanese, and reduces the 'checking...' pauses that interrupt spoken conversations.
- You can try it now by updating your ChatGPT app and opening Voice mode; API access is announced as coming soon, not available yet.
- There is no dedicated product page for this title yet — check openai.com for the official rollout details and any API timing.
Quick Facts
| Type | Voice model update for ChatGPT, Codex and dots |
|---|---|
| Devices | Wherever Voice in ChatGPT is available |
| Price | Included with existing access; check openai.com |
| Languages | New voices: French, German, Hindi, Spanish, Korean, Japanese |
| Highlight | Fewer 'checking...' pauses, better accents |
| Learning curve | None — just talk; pick a voice in settings |
| Category | AI & Models |
| Creator | OpenAI |
| Official site | openai.com ↗ |
| Heat | 🔥 2M Heat |
| Last updated |
What GPT-Live-1.1 Actually Changes in Spoken Conversations
GPT-Live-1.1 is an update to the voice model that powers Voice in ChatGPT, Codex and dots, announced by OpenAI on October 15, 2026. Rather than a brand-new standalone app, it is a refresh of the listening-and-speaking layer you already use when you talk to these products out loud. The headline changes are new voices, better accent accuracy in six languages, and noticeably fewer of the 'checking...' stalls that used to break the rhythm of a spoken exchange.
The language list is specific: French, German, Hindi, Mexican Spanish, Korean and Japanese. That matters because accent quality is where voice assistants usually fall down — a model can understand a language and still sound stilted when it speaks it back. OpenAI frames this release as fixing exactly that, so the assistant sounds closer to a native speaker rather than a textbook reading. The update rolled out to all users on announcement day, with API access described as coming soon.
The 'checking...' reduction is the quieter but arguably bigger change. In a spoken conversation, a two-second silence while the model fetches something feels like a dropped call. Cutting those gaps makes voice mode usable for longer, more natural back-and-forth — brainstorming, hands-free note-taking, or practicing a language — instead of short command-and-response exchanges.
- New voices, not just new languages
The release adds fresh voice options in each of the six languages, so you can pick a tone and accent that fits how you actually speak.
- Accent accuracy is the selling point
Better accents in French, German, Hindi, Mexican Spanish, Korean and Japanese make replies sound less robotic and easier to follow at speed.
- Less dead air
Fewer 'checking...' pauses means the assistant keeps the conversational thread instead of going silent mid-thought.
- Same three surfaces
The update lands in Voice in ChatGPT, Codex and dots at once, so the experience is consistent across those products.
How Voice Mode Works Under the Hood, in Plain Terms
Voice mode is a pipeline: your speech is transcribed, the model reasons about your request, and a voice model speaks the answer back. GPT-Live-1.1 sits at that final stage, and partly at the listening stage, improving how natural the output sounds and how quickly it arrives. When people complain that a voice assistant feels robotic, the culprit is usually this output layer — pacing, intonation and accent — not the underlying reasoning.
Latency is the other half of the equation. Any pause you hear is the model deciding, calling a tool, or waiting on a response before it can speak. The 'checking...' phrase is essentially a verbal placeholder that fills that gap. Reducing how often it appears means the system is either answering faster or better at speaking while it works, which is what makes a conversation feel continuous rather than turn-based.
Because GPT-Live-1.1 is delivered as an update rather than a separate download, you do not manage it directly. You update the app, open Voice mode, and the new behavior is simply there. That also means availability can vary slightly by region and platform as the rollout completes — if you do not see the new voices immediately, check again later or look for guidance on openai.com.
- Transcribe, think, speak
The three-stage pipeline explains why voice quality and response speed are separate problems that this update tackles together.
- Why 'checking...' exists
It is a filler for processing time; hearing it less often signals faster answers or smoother overlap between thinking and talking.
- No separate install
GPT-Live-1.1 arrives through your existing app update, so there is nothing extra to configure beyond choosing a voice.
Trying the New Voices: A Practical Walkthrough
Getting started is deliberately low-effort. Update the ChatGPT app on your phone or device, open Voice mode, and look for the voice or language settings. The new French, German, Hindi, Mexican Spanish, Korean and Japanese voices should appear there once the rollout reaches you. If your app has not updated yet, force a refresh from your app store and reopen it.
Once you have picked a voice, test it the way you would test any new tool: ask something short, then something that requires the model to look something up, and listen for how it handles the gap. That second test is where the reduced 'checking...' behavior shows up most clearly. Try a sentence or two in the target language and judge the accent yourself — your ear is the only benchmark that matters here.
For language learners, the practical trick is to hold a real conversation rather than drill phrases. Ask the assistant to reply only in your target language, correct your pronunciation, and slow down when you ask. Because the update improves accent quality, shadowing the replies — repeating them aloud right after you hear them — becomes a genuinely useful exercise rather than an exercise in imitating a robot.
- Update first, then look in settings
New voices only appear after the app update lands; if they are missing, refresh the app and check openai.com for rollout notes.
- Test with a lookup question
Ask something that requires fetching information — that is the scenario where fewer 'checking...' pauses is most obvious.
- Use it for shadowing practice
Repeating improved-accent replies aloud turns voice mode into a free pronunciation drill for the six supported languages.
- Keep sessions conversational
Longer, natural exchanges reveal latency and pacing issues that short commands never surface.
How It Compares, What It Does Not Do, and Who Benefits Most
Compared with earlier voice experiences in ChatGPT, the pitch here is refinement rather than reinvention: same surfaces, better sound and less waiting. Against other voice assistants, the differentiator is the specific language set — French, German, Hindi, Mexican Spanish, Korean and Japanese — which targets users who have long been served poorly by English-first voice models. If you work or learn in one of those languages, this update is aimed squarely at you.
The limitations are worth stating plainly. API access is announced as coming soon, so developers cannot build on GPT-Live-1.1 yet. There is no dedicated page for this title, meaning specifics like exact voice counts, supported regions and any usage limits are not fully documented — check openai.com rather than relying on secondhand summaries. And an improved accent is not the same as perfect fluency; judge it in your own language before assuming it fits your use case.
The clearest beneficiaries are multilingual professionals, language learners, and anyone who uses voice mode hands-free while commuting, cooking or working. If you mostly type, this update changes little for you. If you talk to ChatGPT, Codex or dots regularly, the combination of better accents and fewer stalls is the kind of upgrade you notice within the first minute of a real conversation.
- Refinement, not a new product
Expect the same Voice mode you know, with better voices and smoother pacing rather than a redesigned interface.
- API users must wait
The API rollout is described as coming soon, so plan integrations around the current model until OpenAI confirms timing.
- Verify specifics officially
Voice counts, regional availability and limits are not detailed in the announcement; treat openai.com as the source of truth.
Step-by-Step Guide
Update your ChatGPT app
Open your app store and install the latest version of ChatGPT. The voice update ships through the normal app update, so an outdated build is the most common reason new voices do not appear.
Open Voice mode
Launch the app and switch into Voice mode using the voice control. This is the same entry point you already use — GPT-Live-1.1 does not add a separate app or menu.
Pick one of the new voices
Go into voice or language settings and select a French, German, Hindi, Mexican Spanish, Korean or Japanese voice. If the list looks unchanged, the rollout may not have reached you yet.
Run a quick latency test
Ask a simple question, then one that needs a lookup, and listen for pauses. Fewer 'checking...' interruptions are the clearest sign the new model is active on your account.
Use it in a real conversation
Hold a multi-turn chat in your target language, asking for corrections and slower speech where needed. This is where better accents and smoother pacing pay off most.
Frequently Asked Questions
What is GPT-Live-1.1 and who made it?
GPT-Live-1.1 is an updated voice model from OpenAI, announced on October 15, 2026. It powers Voice in ChatGPT, Codex and dots, adding new voices and better accents in French, German, Hindi, Mexican Spanish, Korean and Japanese while reducing the 'checking...' pauses that interrupt spoken replies.
How do I try the new voices?
Update your ChatGPT app, open Voice mode, and check the voice or language settings for the new options. The update rolled out to all users on announcement day, though regional timing can vary. If nothing looks different, refresh the app and check openai.com for rollout details.
Is GPT-Live-1.1 available in the API?
Not yet. OpenAI described API access as coming soon when the update was announced, so developers should keep building on the current model for now. Watch openai.com for the official timing rather than assuming a date.
Does it cost extra to use?
The announcement does not describe a separate price for GPT-Live-1.1; it is presented as an update to Voice in ChatGPT, Codex and dots. Any plan-level differences or usage limits are not detailed in the notes, so verify the current terms on openai.com.
Which languages get the improved accents?
The new voices and better accents cover French, German, Hindi, Mexican Spanish, Korean and Japanese. Other languages may still work in voice mode but are not part of this specific update, so test your own language and judge the quality yourself.
