How to make Claude read out loud
Does Claude have text to speech? Where Claude's Read aloud button is, why it stops working, your computer's own voice, and how to make Claude Code talk.
Claude can read out loud, but where depends on which Claude you use. The Claude apps on the web, on your computer and on your phone can read a reply aloud, and voice mode answers you in speech. Claude Code and the Claude API have no text to speech at all: in Claude Code you add a voice yourself, like the one in this demo.
Does Claude have text to speech?
| Where you use Claude | Text to speech |
|---|---|
| claude.ai, the desktop app, the iPhone and Android apps | A Read aloud button on each reply, and voice mode |
| Claude Code, in the terminal or in VS Code | None built in: /voice is dictation, speech to text |
| The Claude API | None: the models return text |
How to make Claude read a reply aloud
In the Claude apps, a reply has a Read aloud button among the buttons under it, with a speaker icon. Click it and Claude reads that reply; while it plays, the button turns into controls to pause, resume and stop. It's there on claude.ai in your browser, in the desktop app for Mac and Windows, and in the iPhone and Android apps.
Anthropic doesn't describe Read aloud in its help center yet, so the button may move or change between app versions. If you don't see it, update the app.
Voice mode: talk to Claude and hear it answer
Voice mode is a spoken conversation rather than a button on one reply. It's in beta on every plan, Free included, on the web, in the desktop app and in the phone apps, and Anthropic says it works best on your phone.
- On the web and on the desktop: click the sound-wave icon at the lower right of the chat, and Stop to leave.
- On your phone: tap the sound-wave icon next to the microphone.
It listens hands-free by default, and you can switch to push to talk. Claude answers out loud in one of a small set of preset voices, and the conversation is saved as text in the chat. Voice conversations count toward your usage limits like typed ones. (Anthropic: use voice mode)
Claude read aloud not working?
The usual reasons, and what to do:
- "You've reached a Read Aloud limit": Read aloud has a usage limit, which Anthropic doesn't publish. Wait and try again later, or use your computer's own voice in the meantime (below).
- "Read Aloud isn't available in this browser": use an up-to-date browser, or the desktop app.
- "Read Aloud is temporarily unavailable": the problem is on Anthropic's side. Try again later.
- No button at all: update the desktop or phone app. Read aloud is new, and older versions don't have it or still have early bugs.
- No sound: check that the browser tab isn't muted and that your computer plays to the speakers or headphones you expect.
Use your computer's own voice
Every computer and phone can read text aloud by itself, with no limit and no account. It reads exactly what's on screen, code blocks and all, in the system voice.
- Mac: turn on Speak selection in System Settings → Accessibility → Read & Speak (on older macOS versions, Spoken Content). Then select Claude's reply and press Option-Esc; press it again to stop. (Apple)
- Windows, in Edge: open claude.ai in Edge, select the reply and right-click → Read aloud selection, or press Ctrl+Shift+U to read the page. (Microsoft)
- Windows, anywhere: Narrator reads whatever you point it at. Start and stop it with Win+Ctrl+Enter.
- Chrome: right-click → Open in reading mode, then press Play. (Google)
- iPhone: turn on Speak Selection in Settings → Accessibility → Read & Speak (Spoken Content on older versions), then select text and tap Speak. (Apple)
- Android: turn on Select to Speak in Settings → Accessibility. Then select text and choose Read aloud, or start Select to Speak with its shortcut and tap the text. (Google)
Does Claude have a text-to-speech API?
No. Claude's models take text and images and return text (Anthropic: models overview), and the OpenAI-compatible endpoint returns no audio. To speak what Claude writes through the API, send the text to a speech service, or to a speech engine on your own computer.
How to make Claude Code read out loud
Claude Code has no text to speech. Its /voice command works the other way round: you speak, and it types your words into the prompt (Anthropic: voice dictation). Its screen reader mode, claude --ax-screen-reader, makes the terminal readable for VoiceOver or NVDA, which then read everything, code included.
That matters more in Claude Code than in a chat, because Claude works on its own for minutes at a time and then waits for you: for an approval to run a command, for an answer, for you to review a plan. If you're not looking at the screen, it waits in silence. You have two ways to make it talk:
- A Stop hook with your system voice. Claude Code's hooks run your own command when a reply finishes, and Anthropic's hooks documentation names read-aloud hooks as a use for it. On a Mac,
jq -r '.last_assistant_message' | sayas a Stop hook reads every reply aloud. It's free, but it reads everything, file paths and code included, and says nothing while Claude waits for an approval. - Verbative Speak, a plugin for Claude Code. When a task finishes you hear a one-line summary written to be heard, and when Claude waits you hear what it needs: an approval, an answer to its question, or a plan to review. Its eleven natural voices run on your computer, with no API key, in any Claude Code session, the terminal included. It's free for 10 spoken summaries a day. Install it in Claude Code:
/plugin marketplace add verbative/verbative-claude-plugins
/plugin install verbative-speak@verbative
Our guide to Claude Code text to speech compares these with text-to-speech MCP servers and skills, step by step, and there's a guide for Windows. If you work in VS Code or Cursor, the Verbative extension also lets you talk back: say your prompt, and answer approvals out loud.
Which one to use
| You want… | Use |
|---|---|
| To hear one of Claude's replies in the app | The Read aloud button |
| A spoken conversation with Claude | Voice mode |
| Any text read aloud, with no limit | Your computer's own voice |
| Claude Code to say what it did and what it needs from you | Verbative Speak |
Frequently asked
Can Claude read out loud?
Yes, in the Claude apps. On claude.ai, in the desktop app and in the phone apps, a reply has a Read aloud button, and voice mode answers you in speech. Claude Code and the Claude API have no text to speech built in.
How do I make Claude read its answer aloud?
Click the Read aloud button with the speaker icon under the reply. While it plays, the same button pauses, resumes and stops it. For a whole spoken conversation, start voice mode with the sound-wave icon instead.
Why is Claude's read aloud not working?
Usually one of three things: you've reached Read aloud's usage limit, your browser isn't supported, or the feature is temporarily unavailable. Wait and try again, switch to an up-to-date browser or the desktop app, or update the app. Your computer's own voice works in the meantime.
Is there a limit on Claude's read aloud?
Yes. When you reach it, Claude says you've reached a Read Aloud limit and to try again later. Anthropic doesn't publish the number.
Can Claude read aloud on the desktop?
Yes. The desktop app for Mac and Windows has the same Read aloud button as claude.ai, and voice mode. Your computer's own voice also works on any text: Option-Esc on a Mac, Narrator or Edge's Read aloud on Windows.
Does Claude Code have text to speech?
No. Claude Code's /voice is dictation, from your speech to text, and its replies stay on screen. To hear them, add a Stop hook that sends each reply to your system voice, or install Verbative Speak, which speaks a short summary when a task finishes and tells you when Claude needs you.
Does Claude have a text-to-speech API?
No. The Claude API returns text. To turn it into speech, pass the text to a speech service or a local speech engine.