Support

Answers to the questions people ask most. If yours is not here, write to contact@halvox.ai.

What you need

A Mac with macOS 14 Sonoma or later, on Apple silicon or Intel, and a microphone. Halvox downloads its speech model once. Models range from 142 MB to 2.9 GB, depending on the one you choose.

Getting started

Open Halvox and follow the short setup. It asks for the permissions it needs and offers to download a speech model.

Then press your hotkey in any app, speak, and press it again. Halvox turns what you said into text and puts it at your cursor.

Halvox also lives in the menu bar, so dictation, file transcription, and your history are one click away.

Permissions

Halvox asks only for what it uses:

• Microphone, so it can hear you.
• Accessibility, so it can paste your words at the cursor in other apps.
• Input Monitoring, so your hotkey works in every app.

Settings, General shows which ones are on. If you declined one, open System Settings, then Privacy & Security, choose the permission, and turn on Halvox. Quit and reopen Halvox afterward.

The hotkey

The hotkey starts and stops dictation from any app. It is Option and Space until you change it.

To change it, open Settings, General, click the hotkey, and press the keys you want. To talk only while you hold the keys down, turn on Hold the hotkey to talk in the same place.

If nothing happens when you press it, check that Input Monitoring is on, then quit and reopen Halvox. If another app uses the same keys, record a different hotkey.

When your words do not appear where you type

Halvox pastes at your cursor when a text field is active and Accessibility is on. Otherwise it copies the text, so you can paste it with Command V.

To leave your clipboard untouched, turn on Use clipboard safe insert in Settings, General. After you dictate, choose Insert Last Halvox Dictation from the Services menu of the app you are typing in.

Languages and accuracy

Halvox detects the language you speak. To stick to one, choose it in Settings, General, Language.

Dictation uses a fast model. For audio and video files, choose a model in Settings, Model: Base is the fastest, Small is balanced, and Large v3 is the most accurate.

If a name or a term keeps coming out wrong, add it in Settings, Dictionary. Each rule replaces what Halvox heard with what you want, on your Mac, every time.

Transcribing audio and video files

Choose Transcribe a file, or drop a file onto the Halvox window. The text appears while Halvox works.

When it finishes, copy the text, export it as plain text or as SRT subtitles, or choose Insert by clicking and then click the text field where it should go.

Cleanup with Claude

Cleanup is optional and stays off until you set it up. It rewrites a dictation in the style you pick, such as a clean paragraph, an email, a Slack message, or bullet points.

It uses your own Anthropic API key. Paste it in Settings, Cleanup. Halvox keeps the key in the macOS Keychain and sends Anthropic only the text of a dictation, never the audio. Anthropic bills that use to your account.

Privacy

Speech recognition runs on your Mac. Your audio and your transcripts stay there.

Choose how long Halvox keeps transcripts in Settings, General: forever, 30 days, or 7 days. The privacy policy has the details.

Siri and Shortcuts

Say “Start dictation with Halvox” or “Stop dictation with Halvox.” In the Shortcuts app, Halvox also offers Transcribe Audio File, which returns the text of an audio or video file.

Contact us

Write to contact@halvox.ai with your Mac model, your macOS version, and what happened just before the problem. Email support, in Settings, Support, starts the message with your Halvox version filled in.