---
title: "ChatGPT Decodes Speech to Text in Real Time: OpenAI Removes Ten-Minute Limit on Voice Input"
description: "OpenAI has improved ChatGPT voice input: speech transcription now appears in real time, and the ten-minute limit on recordings has been lifted. The update was released for all plans on November 19."
date: 2026-11-19T19:39:21.998Z
lang: en
url: https://xab.info/en/posts/chatgpt-voice-input-realtime-transcription-update
tags: [chatgpt, openai, voice-input, speech-to-text, ai]
author: "Alexander Buran"
publisher: "XAB.info"
---

# ChatGPT Decodes Speech to Text in Real Time: OpenAI Removes Ten-Minute Limit on Voice Input

![ChatGPT logo with a black symbol on a gray background for an article about voice input](https://xab.info/media/2026/11/19/chatgpt-voice-input-realtime-transcription-no-time-limit/chatgpt-voice-input-realtime-transcription-no-time-limit-1.webp)

## Key Points

- OpenAI added real-time voice input transcription to ChatGPT, accessible by tapping the sound wave icon.
- The 10-minute limit on voice recordings has been removed; the company claims improved recognition accuracy.
- The update applies to request dictation, not the full two-way AI Voice Mode.
- The feature is available globally on all plans since November 19 for web, desktop, iOS, and Android; the Record mode remains in beta for paid plans.

OpenAI has updated the voice input feature in ChatGPT, adding real-time speech transcription and removing the previous 10-minute limit on voice recordings. According to the company, the update also improves speech recognition accuracy. The changes became available to users worldwide on November 19 in the web version, the desktop app, and on mobile apps for iOS and Android.

### "On-the-Fly" Transcription and Editing Before Sending

Now, while dictating, users can tap the sound wave icon to view the text that ChatGPT is currently transcribing. This allows the user to monitor how the system perceives their speech even before finishing the recording. Once dictation is complete, the resulting transcript can be edited directly in the prompt input field—allowing the user to fix recognition errors or refine wording before sending the text to the AI.

### Removal of the 10-Minute Limit and Increased Reliability

Previously, voice recordings in dictation mode were limited to ten minutes. OpenAI has eliminated this restriction, making the feature suitable for longer spoken requests. The company states that the update not only speeds up speech recognition but also makes it more reliable, reducing the number of errors in the final text.

### Dictation Is Not a Full Voice Mode

It is important to distinguish this updated feature from ChatGPT's full Voice Mode, which allows for two-way conversations with the AI. The changes introduced on November 19 specifically apply to voice dictation—a tool where the user speaks, and the system converts it into an editable text query. This is not the same as a voice dialogue with the model.

### Separate Record Mode and Availability on All Plans

Separate from dictation, OpenAI previously introduced a "Record" mode for transcribing voice notes and meetings. As of this update, it remains in beta and is available only to Plus, Enterprise, Edu, Business, and Pro users. Unlike that feature, the new real-time transcription display and the removal of the time limit are rolling out globally to all ChatGPT tiers, including the free version, and are operational on the web, desktop, and mobile apps.

## Fact-Check & Sources

- [ChatGPT научился расшифровывать речь в текст на лету без ограничения по времени](https://3dnews.ru/1151549) - Primary source: description of the voice input update — real-time, 10-minute limit removal, availability on all plans from November 19, distinction from Voice Mode, and Record mode's beta status.
- [ChatGPT научился расшифровывать речь в текст на лету без ограничения по времени](https://3dnews.ru/1151549/chatgpt-nauchilsya-rasshifrovivat-rech-v-tekst-na-letu-bez-ogranicheniya-po-vremeni) - Duplicate link to the same material; facts align with source 1.
- [ChatGPT научился одновременно говорить и слушать](https://www.newsru.co.il/science_hitech/12jul2026/open_ai_am.html) - Describes a different function — the full-duplex GPT-Live voice model for two-way conversation. Does not contradict the main material; used here to provide context for distinguishing dictation from full voice mode.
- [ChatGPT научился создавать мини-приложения прямо в чатботе](https://www.ferra.ru/news/apps/chatgpt-nauchilsya-sozdavat-mini-prilozheniya-pryamo-v-chatbote-08-10-2026.htm) - Подтверждено по источнику ferra.ru

## FAQ

### What specifically has been improved in ChatGPT voice input?

Real-time speech transcription has been added: while dictating, tapping the sound wave icon reveals the text being transcribed. After recording, the text can be edited in the input field before submission.

### How long can I record a voice request now?

The previous 10-minute limit has been lifted; the duration of voice recording is no longer capped by that threshold.

### Is this a full voice conversation mode with the AI?

No. The update concerns the voice dictation function, not the full voice mode that enables two-way conversations with the AI.

### Which plans and devices support the new feature?

It is available worldwide on all ChatGPT plans, including the free version, as of November 19 on web, desktop, iOS, and Android apps.

### What is the status of the Record mode?

The Record mode for transcribing voice notes and meetings remains in beta and is exclusively available to Plus, Enterprise, Edu, Business, and Pro subscribers.