VoiceOver Audio Split

VoiceOver Audio Split puts VoiceOver’s speech in one ear and everything else in the other, like the audio split in JAWS. It also brings popular features from JAWS, NVDA and SuperNova to the Mac, and every one of them works from the keyboard.

Download VoiceOver Audio Split (installer, 4 MB)

Latest version: 1.2, released November 3, 2026.

What’s new in version 1.2
  • AI on your Mac: image and video descriptions, the assistant, Do It for Me and the money identifier can use Apple's on-device model (macOS 27 with Apple Intelligence): no API key, no cost, and nothing leaves your Mac. Choose in Settings, under Assistant and image descriptions; Claude with your own key is still there for longer answers.
  • Describe images automatically: when the VoiceOver cursor rests on a picture for a second, the AI on this Mac describes it, with no key to press. Great for going through photos. Turn it on in Settings, the Assistant menu, or say "Audio Split, describe images automatically".
  • Record the screen: one command starts and stops a recording of the screen, with the Mac's sound and VoiceOver, and your microphone if you like, saved in Movies, VoiceOver Audio Split Recordings (macOS 15 or later). Give it a key in Settings, under Command keys, or say "Audio Split, start recording".
  • "Audio Split, can you see me?" runs the camera framing check by voice: whether your face is centered, at a good distance and well lit.
  • Resource monitor, like NVDA's Resource Monitor add-on: one command speaks how busy the processor is, memory used and free, space on the startup disk, the battery and how long the Mac has been on. Press 1 to 7 right after it for one part in full: processor (busiest core, whether the Mac is running hot), memory (pressure and swap), every disk, battery health, macOS version and Mac model, network and address, and the apps using the processor most. Each part is a voice command too ("Audio Split, memory", "what's using the processor"), and a window in Everyday Tools shows them all; Return refreshes it, and it only updates by itself if you turn that on, so VoiceOver isn't interrupted. Give the command a key in Settings, under Command keys.
  • Fixed, from a code audit:
  • The hearing profile did nothing unless the split or another sound feature was already on.
  • "Keep this sound setting?" after a preset chosen by hand put back the layout and the other settings but not the split or the balance; and its window was never freed.
  • Spoken yes-or-no questions (restart VoiceOver, switch output, leave the meeting, Do It for Me's press) often heard nothing, because the app's own question was still being discounted when the answer came. A microphone change during a question is handled too.
  • "Audio Split, find a window" opened Do It for Me instead of the window switcher.
  • Do It for Me no longer sends what's typed in text fields (only their labels), always says which control it is about to press, using the app's own name for it, and can't crash on an odd answer from the AI.
  • A picture taken for the AI on this Mac is never sent to Anthropic with a later follow-up unless sending pictures is allowed in Settings.
  • Describing images automatically now cancels when the cursor moves on, and can't describe one picture under another's name.
  • Pressing the recording key again while the movie was being saved reported a good recording as failed; it now says it is still saving. Quitting while recording finishes the movie first.
  • The listening test restarted within nine seconds skipped a tone.
  • Timers between whole minutes were rounded ("a minute and a half" became 2 minutes). Reminders follow a time-zone change without a relaunch.
  • Practice's closing score counted the unanswered task. The Braille window no longer picks up a password manager's copy, translates off the main thread so typing stays smooth, and says when a file had to be uncontracted.
  • In lessons, Command-Tab, Command-Q, Command-W, Command-H and Command-M always work during practice, so you can leave.
  • The setup guide said the AI features need an Anthropic key; they don't on macOS 27. "List Timers and Reminders" and "Cancel the Last Reminder" are in the menu.
  • CI now fails if a new string is missing from the translation catalog, and the system audio permission text is checked and translated again.

SHA-256 of the installer: 2c28c0e9dd448783db1a59f56f0d0f558f9ef852bdcc1ae1ec6235f0ebedd089

Build 202611031940

Open the downloaded installer and follow its steps; the app goes into your Applications folder. It’s signed and notarized by Apple. All versions and release notes.

Or try the beta on TestFlight, which gets new versions first. TestFlight is by invitation: email amena.thomas@fullcircleat.com with the email address of your Apple Account (your Apple ID), and you’ll get an invitation from TestFlight at that address. Install the free TestFlight app from the Mac App Store, then accept the invitation.

It’s free, for macOS 14.2 or later. Headphones are recommended for the split.

Once it’s installed, the app updates itself: it checks GitHub once a day, or choose Check for Updates, and asks before downloading anything.

New in version 1.1

Audio splitting

Control your Mac’s sound

Hearing and clarity

Speech tools inspired by JAWS and NVDA

Reading and low vision, inspired by SuperNova

Audio hub

Internet radio, podcasts, free LibriVox audiobooks and news headlines, all in one keyboard-friendly window.

Everyday tools

Working with apps

Learn as you go

Optional AI

Describe the image under the VoiceOver cursor or what’s happening in a video, ask an assistant how to do something on your Mac, find controls with Do It for Me, or identify money. These features use Claude with your own Anthropic API key. Nothing is sent unless you ask.

Voice commands

Say "Audio Split" and a command, from anywhere: "Audio Split, turn on VoiceOver", "quiet everything", "what time is it?", "read this label", "where am I?". Speech is recognized on your Mac and never recorded; only "find…" sends your request to Anthropic, through Do It for Me. "Launch VoiceOver training" opens the lessons.

In your language

English, Spanish, French and German, following your Mac’s language.

Built for the keyboard

Every command is the right-hand Option key plus another key, so the left Option key keeps working normally. For example:

A setup guide walks you through each permission the first time you open the app.

Privacy

Read the privacy policy

Free and open source

VoiceOver Audio Split is free software under the GNU General Public License version 3. The source code and contributor guide are on GitHub; you don’t need a GitHub account to read them or download the app.

Feedback

Feedback from VoiceOver users shapes what comes next. To report a problem or suggest an idea, email amena.thomas@fullcircleat.com. If you use TestFlight, you can also send feedback from the TestFlight app, and if you have a GitHub account, you can also open an issue on GitHub (needs a GitHub account).

Download VoiceOver Audio Split (installer, 4 MB)

Or try the beta on TestFlight, which gets new versions first. TestFlight is by invitation: email amena.thomas@fullcircleat.com with the email address of your Apple Account (your Apple ID), and you’ll get an invitation from TestFlight at that address. Install the free TestFlight app from the Mac App Store, then accept the invitation.