Voice input in ChatGPT: how the built-in dictation works

Potok transcribes Russian speech; this site has been translated into English for convenience.

ChatGPT has built-in text dictation: the microphone icon in the input field turns on speech recognition, and the resulting text can be edited before sending, as OpenAI's help describes. It is a separate feature from the Voice Mode. You can also dictate text into any field, including ChatGPT, with Potok (potok.now), at about 900 characters per minute by voice.

Voice input in ChatGPT can mean different things: a voice conversation with the model (Voice Mode) or dictation of text into the input field. ChatGPT has a separate built-in dictation; let us see how to turn it on and what can supplement it.

Does ChatGPT have voice input for text

ChatGPT has built-in text dictation: the input field has a microphone icon, and a click or tap on it turns on speech recognition. What you say appears as ordinary editable text in the input field, so you can correct it before sending the message.

This is a separate feature from the ChatGPT Voice Mode; this is how OpenAI's help describes dictation. In the mobile app for iOS and Android it works when microphone access is turned on in the system settings: on iPhone and iPad this is done in Settings, under Privacy & Security → Microphone, where you need to make sure access is turned on for ChatGPT. Details are in OpenAI's help: Voice dictation FAQ.

The ChatGPT text field also accepts ordinary pasting of text, and you can use this if you need a tool other than the built-in microphone for dictation, for example Potok. We look at a similar question, a feature of its own versus a separate voice mode, for Alice too.

How to dictate in ChatGPT with Potok

1. Open ChatGPT in the browser or in the Windows or macOS app and place the cursor in the message input field.
2. Hold Potok's hotkey (Ctrl+Space by default) and, without releasing it, say the message.
3. Release the key: the recording goes to Potok's servers, and the text appears in the same input field.
4. Check the text, fix any recognition inaccuracies if needed, and send the message.

Potok inserts text into the field where the cursor is, so it works for any ChatGPT text field, not only the message line. Punctuation and capital letters are placed automatically, and you do not need to dictate “zapyataya” (comma) or “tochka” (period). Recognition runs in the cloud on Potok's own model trained on Russian speech, including terms and Russian text mixed with English words, which is handy for prompts in two languages. Potok states about 900 characters per minute versus about 200 with ordinary typing. If the default hotkey is inconvenient, you can change it in the settings to a ready combination, your own key or a side mouse button.

Where voice input in ChatGPT with Potok does not fit

If you need to quickly dictate one phrase straight into ChatGPT and install nothing else, the built-in microphone icon in the input field is enough; it is available both in the app and in the browser.

Potok is a cloud service with recognition on servers in Russia; it does not work without the internet. It has no voice commands or voice editing of text: it only inserts the dictated text into the field. It has no offline mode and no phone app: it is a desktop program for Windows, macOS and Linux. Other apps with a similar question are in the section voice input in apps.

Frequently asked questions

Built-in dictation and Potok

Primary sources: chatgpt.com · help.openai.com