# Voice Input

Ask questions about your data without typing a single character. Voice input dictates into the chat composer, so a quick question like "how many rows in the pageviews table?" takes a sentence instead of a keyboard. Nothing is ever sent automatically: your words land in the composer and you decide when to send.

Voice input is available on **Web**, **Desktop (macOS only)**, and **Mobile** (iOS and Android), and each platform uses a different speech engine — which changes the interface you see and the language it transcribes. Read the section for your platform rather than assuming they match.

Note
The Windows desktop app has no voice input, and no microphone button appears there. Use the web app on Windows if you want to dictate.

## Prerequisites

- Treasure AI Studio on Web, Desktop (macOS), or Mobile (iOS or Android)
- **Microphone permission** granted to Treasure AI Studio in your browser or device settings
- On Web, a browser whose speech recognition Studio can use (Chrome, Edge, and Safari provide it; Firefox does not)
- On macOS Desktop, **Dictation** enabled in macOS settings


## Voice Input by Platform

| Platform  | Speech engine  | What you see  | Transcription language  |
|  --- | --- | --- | --- |
| **Web** | Your browser's speech recognition | No separate interface — the transcript streams into the composer as you speak, and the microphone button turns red while listening | English (`en-US`), regardless of the app's language setting |
| **Desktop (macOS)** | macOS system Dictation | macOS's own dictation indicator; the app shows no recording state of its own | The language set for Dictation in macOS |
| **Mobile (iOS and Android)** | The device's speech recognition — iOS speech recognition on iOS, the Android system speech recognizer on Android | A recording panel above the composer with a live waveform, an elapsed timer, and a transcript preview to confirm or discard | The app's language — English or Japanese |


## Dictating on Web

1. Select the **microphone button** in the chat input toolbar. Recognition starts immediately — there is no confirmation step
2. **Speak your message.** The transcript appears in the composer as you talk, refining itself as the recognizer settles on words. If the composer already held text, the transcript is appended after it, separated by a space
3. Select the **microphone button** again to stop listening
4. Edit the text if you need to, then send


Typing or editing the composer by hand while Studio is listening stops recognition, so you can always take over mid-sentence. Sending also stops it.

The microphone button appears only when your browser exposes speech recognition to the page. In a browser without it — Firefox, most notably — the button is absent rather than broken.

Web speech recognition is your browser's, not Studio's
On Web, transcription is performed by the browser's own speech recognition. Depending on the browser, that audio may be processed on your device or sent to the browser vendor's speech service — Studio does not control which, and neither route is covered by Treasure AI's own data handling. If your organization treats spoken input as sensitive, type the message instead.

## Dictating on Desktop (macOS)

Selecting the **microphone button** starts **macOS system Dictation** with the composer focused, so your words go straight into the message you are writing. Because this is the operating system's feature rather than Studio's:

- macOS shows its own dictation indicator; Studio's microphone button does not change state
- You end dictation the macOS way (press the dictation shortcut again, or select **Done** in the macOS indicator) — the app cannot stop it for you
- Dictation transcribes in the language configured for Dictation in macOS, which may differ from the language Studio's interface is set to


Enable Dictation before first use in **System Settings** > **Keyboard** > **Dictation**, and allow microphone access for Treasure AI Studio when macOS asks.

## Dictating on Mobile

1. **Tap the microphone button** in the chat input toolbar
2. A recording panel appears above the composer, which stays visible
3. **Speak your message** — Studio transcribes as you talk
4. **Review the transcript** shown in the recording panel (up to two lines are visible)
5. Tap the **checkmark** to insert the transcript into the composer. If the composer already held text, the transcript is appended after it, separated by a space
6. Tap the **X** to discard the recording and close the recording panel with nothing inserted


While recording, the panel shows an audio waveform that responds to your voice, a pulsing recording dot, and an elapsed timer (for example, "0:05"), so you can tell that Studio is hearing you before you commit the text.

Confirming inserts without sending. You always get to edit before tapping **Send**.

## Language Support

Voice input transcribes in whichever language its platform's engine is set to, and the three platforms resolve that differently:

- **Mobile** (iOS and Android) follows the app's language setting — English or Japanese. Change it in the app before dictating.
- **macOS Desktop** follows the macOS Dictation language, set in **System Settings** > **Keyboard** > **Dictation**.
- **Web** always transcribes as **English (`en-US`)**, whatever the app's language is set to, so Japanese speech does not come out correctly. To dictate Japanese into the web app, use your operating system's own dictation (macOS Dictation, Windows voice typing) with the message box focused — that route follows the language your OS is set to. On a Mac, the desktop app's microphone button already does exactly this.


## Permissions

Voice input requires **microphone access**. The first time you use it, your browser (Web) or device (Desktop, Mobile) prompts you to grant microphone permission to Treasure AI Studio.

If microphone permission was previously denied, re-enable it:

- **Web** — Open the site's permission settings in your browser (in most browsers, from the icon in the address bar), allow the microphone, then reload the page.
- **Desktop (macOS)** — Allow microphone access for Treasure AI Studio in your macOS privacy settings, confirm Dictation is enabled in **System Settings** > **Keyboard** > **Dictation**, then restart the app.
- **iOS** — Open **Settings** > **Treasure AI Studio**, enable **Microphone** and **Speech Recognition**, then return to the app and try voice input again.
- **Android** — Grant microphone access to Treasure AI Studio in your device's app settings, then return to the app and try voice input again.


## Verification

- [ ] Select the microphone and see Studio respond to your voice — the transcript filling the composer on Web, the waveform moving on Mobile, or the macOS dictation indicator on Desktop
- [ ] Confirm the transcript reaches the composer without being sent
- [ ] Edit the transcribed text and send it


## Troubleshooting

| Issue | Solution |
|  --- | --- |
| No microphone button | On the Windows desktop app there is none — voice input is macOS-only on Desktop. On Web, your browser may not provide speech recognition (Firefox does not); use Chrome, Edge, or Safari. On Mobile, update to a recent version of the app |
| Nothing is transcribed | Check that microphone permission is granted for Treasure AI Studio — on iOS in **Settings** > **Treasure AI Studio**, on Android in your device's app settings, on Web and Desktop in your browser or macOS settings. Reload the page (Web) or restart the app (Desktop, Mobile) after changing it |
| Japanese speech comes out as nonsense on Web | Studio's microphone button transcribes English only on Web. Dictate with your operating system's dictation instead (macOS Dictation, Windows voice typing) with the message box focused, or type the message |
| Dictation on macOS never stops | macOS owns the session; end it from the macOS dictation indicator or with your Dictation shortcut. Studio's microphone button cannot stop it |
| Transcript is inaccurate | Speak clearly at a natural pace. Background noise reduces accuracy for every engine — speech recognition works best in quiet environments |
| Dictation replaced text you had typed | It should append after existing text with a space. On Web, editing the composer by hand while listening stops recognition — restart it after you finish typing |


Pro Tip
Combine voice input with typing. Start typing a complex message, then dictate the rest — the transcript is appended to what you already wrote rather than replacing it.

## Next Steps

- [AI Chat Interface](/products/ai-studio/chat/chat) — Full chat interface guide
- [Getting Started](/products/ai-studio/getting-started) — Set up Studio on Web, Desktop, or Mobile
- [Chat History](/products/ai-studio/chat/chat-history) — How conversations are saved across devices