Voice-to-Text

Dictate text into any app. Hold the activation key, speak, and release. The text goes where your cursor is.

Transcription runs on your computer. No audio, transcripts or metadata go to external servers. The only network access is the one-time model download.

Setup

1. Open Preferences → Voice-to-Text and turn on Enable Voice-to-Text.
2. Select a Transcription Approach and a model.
3. Click Download on the model card.

You download a model only one time. The speech engine is part of the app on all platforms. You do not compile or install it. Apple Silicon Macs use the Apple Neural Engine. Linux, Windows and Intel Macs use the included whisper-server.

On macOS, DevWorkspace Pro needs the Accessibility permission to paste or type into other apps. Grant it in System Settings → Privacy & Security → Accessibility when macOS asks.

Record

1. Hold the activation key. The default is right ⌘ on macOS and right Ctrl on Windows and Linux.
2. Speak. An overlay shows at the bottom of your screen.
3. Release the key. DevWorkspace Pro transcribes the audio and puts the text into the focused app.

If DevWorkspace Pro cannot type the text into the focused app, it copies the text to the clipboard. The overlay shows Copied and the reason. Press ⌘V (macOS) or Ctrl+V to paste the text.

Press Esc to cancel and close the overlay.

Recording starts only after you hold the key for the Hold Duration (default 300 ms). A short tap does not start a recording.

If you press another key or click the mouse while you hold the activation key, DevWorkspace Pro treats it as a shortcut (for example, right ⌘ + Enter) and does not record. If the recording already started, it stops without a transcription. A click on the copy icon in the overlay is the exception (see Copy the Last Transcript).

Hands-Free Recording

Double-tap the activation key to record without holding it. Tap the key once more to stop and transcribe. You can type and click during a hands-free recording.

Copy the Last Transcript

If the text went to the wrong app (for example, you changed apps during a long transcription), hold the activation key to show the overlay, then click the copy icon. DevWorkspace Pro stops the recording and puts your last transcript on the clipboard. Press ⌘V (macOS) or Ctrl+V to paste it. DevWorkspace Pro keeps the last transcript until you quit the app.

Limits

• One recording can be up to 10 minutes long. The timer changes color 30 seconds before the limit. At the limit, recording stops and the overlay shows Limit reached · transcribing.
• If your microphone disconnects, recording stops and DevWorkspace Pro transcribes the audio that it has. The overlay shows Mic disconnected · transcribing.

Your computer does not go to sleep during a recording. If the transcription engine stops responding, DevWorkspace Pro restarts it.

Transcription Approach

Approach	Platform	Needs
Automatic	All	Uses the Apple Neural Engine for Neural Engine models on Apple Silicon. Uses whisper.cpp for all other models.
Apple Neural Engine (Core ML)	macOS with Apple Silicon	Nothing. The engine is part of DevWorkspace Pro.
Standard (whisper.cpp)	macOS, Windows, Linux	Nothing. whisper-server is part of DevWorkspace Pro.

Without Apple Silicon, DevWorkspace Pro always uses Standard (whisper.cpp).

Models

Click Show models folder to see the downloaded files.

Apple Neural Engine models (Apple Silicon only):

Model	Language	Note
Parakeet TDT v2	English	Best for English
Parakeet TDT v3	Multilingual	Best multilingual
Whisper Tiny, Base, Small	Multilingual	Under More WhisperKit models
Whisper Large v3	Multilingual	Most accurate

Standard (whisper.cpp) models:

Model	Language	Note
small.en	English	Recommended default
tiny.en	English	Fastest
base.en, medium.en	English
large-v3	Multilingual	Most accurate
tiny, base, small, medium	Multilingual	Under More models

An English-only (.en) model is more accurate than a multilingual model of the same size. For other languages, use a multilingual model and set Recognition Language.

Settings

Setting	What it does
Activation Key	The key that you hold to record.
Hold Duration	How long you must hold the key before recording starts
Double-tap for hands-free	On by default. Turns hands-free recording on or off.
Insert via clipboard	On by default. Puts the text on the clipboard and pastes it with ⌘V (macOS) or Ctrl+V. This works well in terminals and code editors. When off, DevWorkspace Pro types each character one at a time.
Keep transcript on clipboard	Keeps the text on the clipboard after the paste. When off, DevWorkspace Pro restores your previous clipboard contents. If you copy something new right after the paste, DevWorkspace Pro keeps your new copy.
Recognition Language	The language that you speak. Applies only to multilingual models.

Linux

DevWorkspace Pro uses a tool to type the text into the focused app:

• X11: xdotool.
• Wayland: ydotool. The ydotoold service must run.

If the tool is missing or fails, DevWorkspace Pro copies the text to the clipboard, and the overlay tells you why. Press Ctrl+V to paste the text.

To install xdotool, run the command for your distribution:

# Debian, Ubuntu
sudo apt install xdotool

# Fedora
sudo dnf install xdotool

# Arch
sudo pacman -S xdotool

You do not have to restart the app. The next dictation types the text automatically.

Wayland limits

• The activation key works only when an X11 or XWayland window has focus.
• To type into native Wayland apps, install ydotool and start the ydotoold service.

Platform Support

Voice-to-Text runs on macOS, Windows and Linux. It is fully tested only on macOS. To report a problem on another platform, open an issue in the community repository.