Guides

Dictate prompts to ChatGPT, Gemini and other AI chats on a Mac, by voice

By · Published · Updated

To dictate a prompt to ChatGPT, Gemini or another AI chat on a Mac, click into the chat's message box, in the desktop app or in the browser, start a dictation tool, say the prompt, and send it once it reads right. Apple Dictation works there, some AI apps have their own voice input, and YapToText, the free app I built, types the prompt at your cursor in any of them.

With YapToText, your voice is turned into text on your Mac, not on a server. The prompt itself still goes to whichever AI service you send it to, exactly as a typed one would, and this page is plain about that. My hands make typing difficult, so I say my prompts every day; below are the steps, the mode to use, and what to do when a prompt does not land.

Download YapToText on the Mac App StoreDownload for free. No account, nothing locked, macOS 14 or later on Apple silicon. Source on GitHub.

On this page: 11 sections
  1. Steps
  2. Why say a prompt instead of typing it
  3. Know where your words go
  4. Dictate a prompt with YapToText
    1. Step 1: Click into the message box
    2. Step 2: Tap Right Command and say the prompt
    3. Step 3: Tap it again
    4. Step 4: Read it, then send
  5. Use Raw Transcription or Clean Up for prompts
  6. Check it, fix it, then send it
    1. Look before it lands
    2. Send it the moment it lands
    3. Fix one word
    4. Teach it your names
  7. Compare the AI apps' own voice input
  8. If the prompt does not land
  9. What it cannot do
  10. The short version
  11. Questions

Steps

  1. Click into the message box of ChatGPT, Gemini or another AI chat, in the desktop app or the browser.
  2. Tap Right Command and say the whole prompt; press 2 for Raw Transcription or 3 for Clean Up while you talk if you want to choose the mode.
  3. Tap Right Command again, and the prompt is pasted at the cursor.
  4. Read the prompt and fix a word with replace, scratch that or Quick Edit.
  5. Send it, or let After a dictation lands press Return for you in a desktop app.

Why say a prompt instead of typing it

A good prompt is long. It has the background, what you already tried, what you want back and what to leave out, and every one of those is a sentence you could say out loud in a few seconds. A prompt is ordinary language, and ordinary language is what dictation is good at.

My hands make typing difficult, so I say my prompts, and these are the parts that dictate well:

  • The background. Who the answer is for, what the project is, what happened last time.
  • The constraints. How long, what tone, what format, what not to do.
  • The follow-up. "That is close, but the second paragraph is too formal."

The parts that do not dictate well are the parts that are not words: code, exact symbols and strings that must be spelled one character at a time. The guide to dictation for coding on a Mac covers that split, and the advice there holds here too: if a person or an AI will read it, say it; if it has to be exact, type it or paste it.

Know where your words go

Dictating a prompt involves two things: your voice, and the text it becomes. They do not go to the same place.

  • Your voice, with YapToText. Whisper Large v3 Turbo (Q5) turns it into text on your Mac, and any cleanup runs on Phi-3.5 Mini Instruct, also on your Mac. There is no account, no analytics, and the only network use is a model download you start yourself.
  • Your prompt, once you send it. It goes to whichever service you are talking to: OpenAI for ChatGPT, Google for Gemini. That is true of a typed prompt too, and their own terms decide what happens next. YapToText cannot change that.
  • Your voice, with the apps' own voice input. OpenAI's Voice Dictation FAQ says the audio you record in ChatGPT is sent to OpenAI's models to be transcribed. For Apple Dictation, the text below Dictation in System Settings, Keyboard says whether your voice is processed on your Mac or sent to Siri servers.

YapToText keeps your voice on your Mac; the prompt you send still goes to the AI service, the same as anything you type there.

Dictate a prompt with YapToText

YapToText is the free Mac app I built for dictating into every app. It types at the cursor, so it works the same in the ChatGPT or Gemini app as in a browser tab with either of them open. It needs an Apple silicon Mac with macOS 14 or later, and it is free on the Mac App Store. Click into the message box, tap, talk, tap, and the prompt is in the box, waiting for you to send it.

Step 1: Click into the message box

  • You: Click into the chat's message box, in the desktop app or in the browser tab.
  • YapToText: Nothing yet. It types into the app in front, never into its own window.
  • Check: The cursor is blinking in the message box, not in the browser's address bar or a search field.

Step 2: Tap Right Command and say the prompt

  • You: Tap Right Command once and say the whole prompt, the way you would explain it to a person. Pause to think whenever you need to.
  • YapToText: Opens the floating panel with your voice as a wave and your words as they form. Auto-stop after silence is off by default, so a pause does not end the dictation.
  • Check: The wave moves when you speak. If it stays flat, the app is not hearing your microphone yet.

Step 3: Tap it again

  • You: Tap Right Command again.
  • YapToText: Finishes the transcript on your Mac, pastes it at the cursor and puts back whatever you had copied before.
  • Check: The prompt is in the message box. If it went to the clipboard instead, Accessibility is not on yet; the fixes further down cover it.

Step 4: Read it, then send

  • You: Read the prompt once, fix anything misheard, and send it the way you always do.
  • YapToText: Nothing more, unless you turned on After a dictation lands for this app, which presses the key for you.
  • Check: The prompt says what you meant.

Long prompts are fine: there is no word cap. While you talk, Space pauses and Esc cancels, and a canceled dictation is still saved to History. The first dictation guide covers the panel and the permissions.

YapToText's Dictation settings page on a Mac: the Keys, Recording, Microphone and Delivery cards.
Figure 1. On the Dictation page, Auto-stop after silence and Maximum length are both Off by default, so a long prompt with pauses in it is not cut short.

Use Raw Transcription or Clean Up for prompts

The mode decides what happens to your words after Whisper hears them. For prompts, two modes fit best. Clean Up drops the ums and keeps your wording; Raw Transcription types exactly what you said.

  • Clean Up. Removes filler words and false starts, fixes grammar and punctuation, and repairs words the microphone clearly misheard. It never deletes, shortens or summarizes a sentence, writes spoken web and email addresses in proper form, and turns spoken numbers into digits. That suits a prompt you thought through out loud.
  • Raw Transcription. Literal speech to text with basic punctuation, and no AI at all. That suits a prompt where every word was deliberate.
  • Not Code Comment. It compresses what you say into short imperative text, which suits a commit message better than a prompt full of context.

Auto mode, on by default, picks for you, and the ChatGPT and Gemini apps are not on any of its app lists, so they get no lean. Neither does a browser. A prompt of ordinary prose gets a careful pass that fixes what was misheard and rewrites nothing, and six words or fewer stay exactly as spoken. Two things are worth knowing, because a prompt is often an instruction:

  • Instructions at the end are Auto's. A prompt that ends with "make it shorter", "make it formal" or "as a bullet list" is read as an instruction to YapToText. The phrase is not typed, and Auto rewrites your prompt that way instead.
  • A letter becomes an email. A dictation that opens with a greeting and closes with a sign-off comes out as an email body. That is what you want in Gmail, and maybe not when you are dictating a draft for the AI to review.

Say: A prompt for a chat app, while Auto mode is on.

rewrite my cover letter so it sounds less stiff, make it shorter

What happens: Auto takes "make it shorter" off the end and shortens your prompt itself, so the AI never sees that part. Press 2 while you talk, for Raw Transcription in the default order, and the whole sentence is typed as you said it.

Three ways to choose the mode yourself:

  • For one dictation. While you talk, press 2 for Raw Transcription or 3 for Clean Up, in the default order. With Auto on, the next dictation goes back to Auto.
  • For one app, every time. Turn Auto mode off, open Clean Up's editor on the Modes page, and add the ChatGPT or Gemini app under Apps that use this mode. While Auto is on, it decides instead, so this takes over only with Auto off.
  • From the menu bar. Click the capybara and pick a mode in the Mode picker. That turns Auto off until you switch it back on.
The Switch Mode panel listing Auto, Raw Transcription, Clean Up, Note, Email, Message and Code Comment, numbered 1 to 7, each tagged AI cleanup or Verbatim.
Figure 2. The number beside each mode is the key that picks it while you talk: 2 for Raw Transcription and 3 for Clean Up, in the default order.

The modes guide explains each mode, and the Auto mode guide lists every phrase Auto follows.

Check it, fix it, then send it

A wrong name or number in a prompt wastes the answer. Read the prompt once before it goes, or let YapToText send it for you where a typo costs nothing.

Look before it lands

Review before inserting, in Settings, Advanced, Text output, holds the finished text in a small editor first: Command Return inserts it and Esc discards it. It is off by default, and while it is on, Only for long dictations (200+ characters) keeps short ones out of it unless you switch that off.

Send it the moment it lands

After a dictation lands, in Settings, Advanced, Text output, can press Return or Command Return in an app the moment your words arrive. Choose Add App, pick the app, and choose the key. In a chat where Return sends the message, that saves a key press. It presses the key without a second look, so anything Whisper misheard goes out too. It is set per app, not per tab, so I would expect turning it on for Safari or Chrome to press Return after every dictation in every tab, Gmail and forms included. I have not tested that. I would only turn it on for a desktop AI app, not for a browser.

Fix one word

  • Scratch that. Right after the words land, dictate "scratch that" on its own, within 30 seconds, and the dictation is taken back. "Replace", the wrong word, "with", the right word swaps one word in place. Both need Insert text automatically, which is on by default.
  • Quick Edit. Select a word or the whole prompt, tap Right Option, say the change and tap again. "Change this to Postgres" swaps the selection for exactly those words, with no AI involved, and "make this shorter" rewrites it on your Mac. The result replaces your selection.

Teach it your names

Prompts are full of names: people, products, libraries, places. On the Dictionaries page, an entry says when you hear this, write that, and the same entries also prime Whisper before it decodes your audio, so the name has a better chance of coming out right in the first place.

Say: Or skip the Dictionaries page: select the right spelling anywhere, tap the Quick Edit key, and say this.

add this to my dictionary

What happens: The card says Added, with the word in quotes, and the entry is in place at once.

YapToText's Quick Edit page: the Quick Edit key on Right Option, Corrections by voice, pop-up position and colors.
Figure 3. The Quick Edit page: Corrections by voice handles scratch that and replace right after the words land, and the Quick Edit key, Right Option, rewrites or swaps a selection.

The Quick Edit guide lists what you can ask for, and the dictionaries guide covers Heard, Replace with and the options.

Compare the AI apps' own voice input

Some AI apps have their own voice input. Here is what each vendor's help pages say. I have not tested these features myself, so this is their description, not my review.

  • ChatGPT. OpenAI's help center separates Dictation, for when you want to record a prompt, review and edit its transcription, then send it as text, from Voice, which is for live back-and-forth conversation. Its Voice Dictation FAQ says the recorded audio is sent to OpenAI's models to be transcribed, and that you can edit the text before you send it.
  • Gemini. Google's help page for the Gemini app on Mac says to click the Mic to say your prompt, then click Send. The same app has Speak to Window, which dictates into whatever window you are using; the Google voice typing on a Mac page covers it.

Checked October 5, 2026, on OpenAI's, Google's and Apple's help pages. Features change, so check before you rely on one.

An AI app's own voice inputApple DictationYapToText
Where it worksInside that app; Gemini's Speak to Window also types into other windowsAnywhere you can typeAny app you can type in, browsers included
Where your voice goesSet by that vendor; OpenAI says ChatGPT's dictation audio goes to its modelsOn the Mac or to Siri servers; Keyboard settings says whichStays on your Mac
Where the prompt goesTo that AI serviceTo whichever AI service you send it toTo whichever AI service you send it to
Silence timeoutSee the vendor's helpStops after 30 seconds of silenceNone by default
CleanupSet by that vendorAdds commas, periods and question marks in supported languagesRaw Transcription, Clean Up or Auto, on your Mac
CostPart of that appFree, built into macOSFree, no paid tier

Apple says Dictation works anywhere you can type, so it works in every one of these chats, with one catch for long prompts: it stops by itself after 30 seconds without speech, which can end a prompt while you think. The 30-second limit page explains why. Use an app's own voice input if you live in one app; use one key that works in all of them if you do not.

If the prompt does not land

Most of the time it is the window or the permission. A prompt that lands in the wrong place is still in History. Climb this ladder, smallest step first:

  1. Click into the message box first. YapToText never types into its own window, so a dictation started with YapToText in front goes to the clipboard. Press Command V and the prompt is there.
  2. Turn on Accessibility. On YapToText's Home page, press Grant next to Accessibility and switch the app on in System Settings, Privacy & Security, Accessibility. Without it, every dictation is copied to the clipboard instead of typed.
  3. Check what was selected. Dictating over selected text replaces it. If part of an earlier prompt was highlighted, click once in the box to clear the selection before you start.
  4. Stop a beep. Intelligent Insert reads a few words around your cursor with invisible keystrokes, and some apps answer one with the alert sound. On the Dictation page, in the Delivery card, choose Turn off in another app and pick that app, or lower Alert volume in System Settings, Sound. The 1.5.2 release notes say Intelligent Insert is more reliable in Gmail, chat pages and Electron apps.
  5. Pin a method. If an app rejects pasted text, open Settings, Advanced, choose Add App Override under Per-app method, pick the app and choose Type character-by-character.
  6. Find it in History. By default History keeps your last 500 dictations for 30 days, so a long prompt lost to a reloaded page can be copied or inserted again.
YapToText's History page: a search field, date and mode filters, and dictations tagged with their mode, app and time.
Figure 4. History keeps each dictation with its mode and the app it went to, so a prompt lost to a closed tab can be copied or inserted again.

The History guide, the Intelligent Insert guide and the troubleshooting guide go further.

What it cannot do

  • Keep your prompt private from the AI. Your voice stays on your Mac. The prompt you send goes to the AI service, like typed text does.
  • Hold a conversation. YapToText types text. It does not read the answer aloud or talk back; a vendor's own voice mode is built for spoken back-and-forth.
  • Know which site is in a tab. Auto mode and After a dictation lands see the browser as one app, so a setting for Safari applies to every page in Safari.
  • Run on older Macs. It needs Apple silicon and macOS 14 or later, and nothing is claimed for Intel Macs.

For what on-device AI does after Whisper hears you, AI dictation on a Mac has the bigger picture.

The short version

  • Say it. Click into the chat, tap Right Command, say the whole prompt, tap again.
  • Pick the shape. Raw Transcription or Clean Up for prompts, and press 2 when a prompt ends in "make it shorter" or "as a bullet list".
  • Check, then send. Review before inserting to look first, After a dictation lands to send by itself in a desktop app, Quick Edit to fix a word.
  • Know where it goes. Your voice stays on your Mac; the prompt goes to the AI you send it to.

Say the prompt the way you would explain it to a person, and read it once before it goes.

If an AI app or a chat page does not take the words, tell me which one in an issue on GitHub. Every one gets read.

Download YapToText on the Mac App StoreDownload for free. No account, nothing locked, macOS 14 or later on Apple silicon. Source on GitHub.

About the author. Ryleigh Newman is a developer whose hands make typing difficult and who dictates every day with YapToText, the app this site describes. Facts about other apps come from their own sites, with the date they were read. More about the project.

Sources

Questions

Can I use voice to text with ChatGPT on a Mac?

Yes. OpenAI's help center describes Dictation in ChatGPT for recording a prompt, editing the transcription and sending it as text. Apple Dictation and YapToText also type into ChatGPT's message box, in the app or in a browser.

Does dictating a prompt keep it private?

Only the voice part. YapToText turns your voice into text on your Mac, but the prompt you send goes to the AI service, the same as a typed prompt. OpenAI says audio recorded with ChatGPT's own dictation is sent to its models to be transcribed.

Which YapToText mode is best for AI prompts?

Clean Up, which drops filler and keeps your wording, or Raw Transcription, which types exactly what you said. Avoid Code Comment for prompts, because it compresses your words into short commit-style text.

Why did the end of my prompt disappear?

With Auto mode on, a dictation that ends with a phrase like make it shorter or as a bullet list is read as an instruction to YapToText, so it is not typed. Press 2 for Raw Transcription while you talk and the whole prompt is typed as said.

Can YapToText send the prompt for me?

Yes. In Settings, Advanced, Text output, add the app under After a dictation lands and choose Press Return or Press Command Return. It presses the key without a second look, and in a browser it applies to every tab.