Voice

Voice lets you work by saying short commands, such as “next segment”, “translate segment” or “undo”. A command can do something in Workbench, or press keys in whichever program is in front: Trados, memoQ, Word or anything else. Voice runs offline, on your own computer.

Voice is for commands only. It never types what you say.

The Voice tab, with the Edit Voice Command dialog open

Switching Voice on and off

Voice listens until you switch it off. There are three ways to switch it on or off:

the ▶️ Start Always-On / ⏹️ Stop Always-On button on the 🎤 Voice tab;
Ctrl+Alt+O, from any program (⌘⌥O on a Mac);
the grey microphone icon in the Windows system tray: click it, or right-click it and choose ▶ Start Always-On or ⏹ Stop Always-On. The icon turns red while Voice is listening. Its menu also has 🎤 Open Voice, which opens the Voice tab.

The line next to the button shows what Voice is doing:

Status	Meaning
⚪ Not active	Voice is off.
⏳ Starting…	Voice is loading its speech model. The first time, it downloads it first (about 40 MB).
🟢 Listening for speech…	Voice is on and waiting for you to speak.
🔴 Recording…	Voice hears you speaking.
⏳ Processing…	Voice is working out what you said.

If Voice stops by itself, for example because the microphone was unplugged, the status goes back to ⚪ Not active. Click ▶️ Start Always-On to start it again.

What Voice understands

Commands are spoken in English, whatever the languages of your project.
Voice only listens for the phrases in your command list. Anything else is ignored and never typed: the status bar just says No command matched. So you can leave Voice on while you talk to a colleague or take a phone call.
Commands go to the program in front. A command that presses keys, such as “undo”, presses them in whichever program you are working in. Click into Trados, memoQ or Word before you speak. Commands that run a Workbench action, such as “translate segment”, always act in Workbench.

How a phrase is matched

A command fires when you say its phrase, or one of its aliases, on its own. Capitals don’t matter.
A phrase of two or more words also fires when Voice hears it inside a longer sentence.
A one-word phrase fires only when you say it on its own, never in the middle of a sentence.
Voice never guesses. If what you said is close to a command but not the same, nothing happens.

What I heard

What I heard, under the on/off button, lists the last 50 things Voice heard, newest first, and what happened:

green: a command ran, for example “undo” → ✓ Undo last action;
red: a command was recognised but failed, for example because a key couldn’t be sent;
grey: no command matched. [unk] stands for words that aren’t in your command list.

When a command didn’t do what you expected, look here first. The list is cleared when Workbench closes.

The built-in commands

These are the commands Voice starts with. You can change, switch off or remove any of them, and 🔄 Reset brings back this list.

Say	Or	What it does
next segment		Go to the next segment
previous segment		Go to the previous segment
first segment	go to start	Go to the first segment
last segment	go to end	Go to the last segment
confirm	confirm segment	Confirm the segment and go to the next unconfirmed one
copy source	copy from source, source to target	Copy the source text to the target
clear target	delete target	Clear the target text
undo		Press Ctrl+Z
redo		Press Ctrl+Y
translate segment	translate this	Translate the segment with AI
translate all segments	batch translate	Translate all segments
lookup	super lookup	Open SuperLookup
concordance	search memory, search TM	Open the concordance search
save project		Press Ctrl+S
open project		Press Ctrl+O
show log	open log	Open the Log window
show editor	editor tab, go to editor	Go to the Editor tab
stop listening		Switch Always-On off, like the ⏹️ Stop Always-On button (from v1.10.375; before that it did nothing)
termbase	add term, add to termbase	memoQ: add a term pair to the termbase (Alt+Down)
tag next	next tag, insert tag	memoQ: go to the end of the segment, insert the next tag and confirm
confirm memoQ	confirm memo	memoQ: confirm the segment (Ctrl+Enter)
confirm trados	confirm studio	Trados Studio: confirm the segment (Ctrl+Enter)

“undo”, “redo”, “save project” and “open project” press keys, so they act in the program in front. The memoQ and Trados commands need AutoHotkey (see below).

If you used Voice before v1.10.374, your own list was kept as it was. Only “start dictation” was removed.

Your own commands

The Voice Commands table on the right of the Voice tab lists all your commands:

Column	What it shows
☑	Whether the command is on
Phrase	What you say
Aliases	Other phrases that do the same
Type	Command, Keystroke, AHK Inline or AHK Script
Action	The command’s description, or what it does if it has none
Category	A label to group commands, such as editing or memoq

Click a column heading to sort the table by that column.

Adding, changing and removing commands

To add a command, click ➕ Add.
To change one, double-click it, or select it and click ✏️ Edit.
To remove one, select it and click 🗑️ Remove.
🔄 Reset replaces your list with the built-in commands.

In the Add Voice Command or Edit Voice Command box:

Phrase: what you say.
Aliases: other phrases for the same command, separated by commas.
Type: what the command does.
- Internal Action (Supervertaler) runs a Workbench action. Pick it under Preset.
- Keystroke presses keys in the program in front. Click the Keystroke field and press the keys, for example Ctrl+Enter. It doesn’t need AutoHotkey. A keystroke recorded as Ctrl+S on Windows is sent as ⌘S on a Mac.
- AutoHotkey Code runs a few lines of AutoHotkey v2 code, for example Send "^{Enter}". Windows only.
- AutoHotkey Script File runs an .ahk file, given by its full path. Windows only.
Description: shown in the table’s Action column.
Category: pick one, or type your own.

The box explains what to fill in for the type you chose.

Changes take effect straight away, also while Voice is listening.

Switching commands on and off

A command that is switched off is greyed out and never fires, but keeps its settings.

One command: click its tick box.
All commands: click the tick-box column’s heading. This switches them all on, or all off if they are all on already.
Several commands: select them with Ctrl+click or Shift+click, right-click and choose ✅ Activate or ⬜ Deactivate.

Microphone and sensitivity

These are on the left of the Voice tab, and are saved as soon as you change them.

Microphone: the microphone Voice listens to. System default uses the one selected in Windows. If the microphone you chose is unplugged, Voice uses the system default.
Mic sensitivity: how loud a sound must be to count as speech.
- Low (noisy): ignores quiet background noise, but may miss soft speech.
- Medium (normal): the default, for a typical home office.
- High (quiet): picks up quiet voices, but may react to background noise.

AutoHotkey

AutoHotkey v2 is only needed for the AHK Inline and AHK Script types, including the built-in memoQ and Trados commands. Under the commands table, Voice shows ✅ AutoHotkey detected or ⚠️ AutoHotkey not found.

📂 Scripts folder opens the workbench\voice_scripts folder in your data folder, a handy place to keep your .ahk files.

The on/off hotkey

The hotkey is Ctrl+Alt+O (⌘⌥O on a Mac). It works from any program. To use another key, click Change in Settings → Keyboard Shortcuts on the Voice tab and change Voice Always-On (toggle). Any key works, for example numpad +.

AltGr counts as Ctrl+Alt on Windows. If AltGr+O types a character on your keyboard layout, you get the character and not the hotkey. Use the left Ctrl and Alt keys instead.
On a Mac, global hotkeys need Accessibility permission. See Keyboard Shortcuts.

Where Voice keeps its files

All in your data folder:

your commands: workbench\settings\voice_commands.json;
your AutoHotkey scripts: workbench\voice_scripts\;
the English speech model: vosk-models\.

🤖 These docs are AI-readable: llms.txt · full · Workbench only