ChatGPT Live Dictation drops the 10-minute limit on all plans

OpenAI has started rolling out Live Dictation in ChatGPT, a new version of the microphone input that turns speech into text while the user is still talking and removes the 10-minute recording limit that applied to dictation until now. The company announced the change on November 18, 2026, in its ChatGPT release notes and in posts on its official X accounts, where it described dictation in ChatGPT as "getting faster and more reliable."
The update reaches every ChatGPT plan, from Free to the paid consumer and business tiers, and it covers the iOS and Android apps, the web version and the desktop app. OpenAI said the rollout is global and gradual, so some accounts will see the new behavior later than others. Users start dictation exactly as before, with the microphone icon in the message box.
Two things change in practice. The first is the waiting time: with the previous system ChatGPT recorded the whole clip and only then produced the transcript, while Live Dictation shows the words on screen as they are spoken, and a tap or click on the waveform shows or hides the running text. The second is length: the old cap stopped a recording at ten minutes, and OpenAI has removed it without announcing a new maximum duration.
··········
CHATGPT LIVE DICTATION: WHAT CHANGED COMPARED WITH THE OLD DICTATION.
Dictation is the feature that converts a spoken message into editable text in the ChatGPT composer. It is a separate function from ChatGPT Voice: the user still reads, corrects and sends the transcript as a normal typed message, and ChatGPT answers in text. OpenAI's help documentation describes the flow in these terms, with the audio sent to OpenAI's models for transcription and the resulting text available for editing before it is sent.
The table below sets the behavior before the November 18 update against the behavior OpenAI describes for Live Dictation.
........
Item | Before November 18, 2026 | With Live Dictation |
Recording length | Capped at 10 minutes | No 10-minute cap; no new maximum announced |
When text appears | After the recording ends | While the user is speaking |
Running transcript | Not available | Shown or hidden by tapping the waveform |
Plans | All ChatGPT plans | All ChatGPT plans |
Platforms | iOS, Android, web, desktop | iOS, Android, web, desktop |
Editing before sending | Yes | Yes |
........
The new release builds on earlier changes to the same feature. On June 26, 2026, OpenAI rolled out a new speech-to-text model for dictation across all plans, and in August it added an automatic retry on iOS when dictation failed because of a connection issue or timeout, plus an Android change that kept the dictation button available after text or attachments had been added. The November update is the first of these to change how long a user can speak and when the text becomes visible. OpenAI has not said which transcription model powers Live Dictation, and it has not published latency or accuracy figures for it.
··········
CHATGPT LIVE DICTATION AVAILABILITY ON FREE, PLUS, PRO AND BUSINESS PLANS.
OpenAI made no distinction between plans for this feature. Free users get the same Live Dictation behavior as subscribers, and the removal of the 10-minute cap applies to all tiers. Readers who want to see how this fits with the rest of the free tier can check our guide to the ChatGPT Free plan and what it includes, while the caps that still differ by subscription are listed in our overview of ChatGPT usage limits by plan.
........
Question | Answer from OpenAI |
Who gets it | Every ChatGPT plan, including Free |
Where | iOS, Android, web and desktop app |
Regions | Global rollout |
Timing | Rollout began November 18, 2026, gradually by account |
How to start | Microphone icon in the message box |
Recording cap | 10-minute limit removed |
Cost | No extra charge announced |
........
For people who already dictate long prompts, the immediate gain is that a long briefing, a draft email spoken while walking, or notes from a call no longer have to be split into ten-minute pieces and pasted together. Seeing the transcript while speaking also makes it easier to catch a misheard name or figure before the message is sent, instead of rereading a long block of text at the end. Accounts that still show the old behavior have not yet received the rollout, since OpenAI said the feature may take time to appear. Users who run into recognition or connection problems can check our troubleshooting guide for ChatGPT voice transcription errors.
··········
[[ADPLUS_300x250_1]]
CHATGPT DICTATION, CHATGPT VOICE AND CHATGPT RECORD COMPARED.
ChatGPT now has three different ways to turn speech into something the model can use, and the November change makes the line between them less obvious. Dictation produces a text message from the user. ChatGPT Voice, rebuilt on GPT-Live since July 8, 2026, is a spoken two-way conversation, and OpenAI is retiring the older Advanced and Standard voice on December 11, 2026, as covered in our report on the move to GPT-Live. Record is a meeting and note-taking tool that transcribes long sessions and produces summaries.
........
Feature | What it produces | Plans | Platforms | Length limit |
Live Dictation | Editable text message | All plans | iOS, Android, web, desktop | No 10-minute cap |
ChatGPT Voice (GPT-Live) | Spoken conversation | ChatGPT plans, with plan usage limits | ChatGPT apps and web | Set by plan usage limits |
ChatGPT Record | Transcript and summary canvas | Plus, Pro, Business, Enterprise, Edu | macOS desktop app only | 4 hours per session |
........
The comparison shows where Live Dictation now sits. Record still offers features that dictation does not, such as speaker labels and automatic summaries saved as canvases, and it remains limited to paid workspaces on the macOS app with a 240-minute cap per session. Live Dictation is the only one of the three that is available on the Free plan across every platform with no published time limit, which makes it the simplest option for anyone who wants long spoken input turned into a prompt.
··········
CHATGPT DICTATION AUDIO: TRAINING, RETENTION AND PRIVACY SETTINGS.
Longer recordings mean more audio passing through OpenAI's systems, so the existing data rules carry more weight than before. OpenAI's Voice Dictation FAQ states that if a consumer user has chosen to share audio to improve its models, OpenAI may train on audio from dictation. Business users are not trained on by default, and OpenAI refers organizations to its enterprise privacy documentation for the handling of business data.
According to the same help page, dictation audio is kept as long as the chat it belongs to stays in the user's history. When the chat is deleted, the associated audio is deleted within 30 days, except when OpenAI must keep it for security or legal reasons or when it was previously shared for training and disassociated from the account. The page also notes that recent legal developments may affect retention.
Consumer users can limit this in the data controls section of ChatGPT settings. Turning off the "Include your audio recordings" toggle stops dictation audio from being shared, and turning off "Improve the model for everyone" stops sharing all audio from dictation and voice chat. Users on Free, Plus and Pro who plan to dictate long personal or work material should review these two settings before relying on the new behavior, because OpenAI's announcement did not change any of them.
··········
CHATGPT LIVE DICTATION: CONFIRMED FACTS AND OPEN QUESTIONS.
OpenAI has confirmed the core of the change: transcription now appears while the user speaks, the running text can be shown or hidden from the waveform, the 10-minute recording cap is gone, and the feature is rolling out globally to all ChatGPT plans on iOS, Android, web and desktop from November 18, 2026.
Several points remain undocumented. OpenAI has not stated whether a new maximum recording length exists, how very long recordings behave on unstable connections, which model handles the live transcription, or how accuracy compares with the June 2026 dictation model. It has also not given a date by which the rollout will reach every account. The data rules that apply are the existing ones in OpenAI's dictation FAQ, which allow training on consumer dictation audio only when the user shares audio for model improvement and keep business accounts out of training by default. For users, the measurable change is the end of the ten-minute ceiling on every plan, including Free.
··········
FOLLOW US FOR MORE.
·····
DATA STUDIOS
·····
datastudios.org




