Voice Typing: dictate into any app on macOS, Windows & Linux
Voice Typing is built into the Colabio desktop app and works system-wide: your words are typed into whatever application has focus — email, docs, chat, your IDE. There are three ways to dictate: hold-to-talk (hold a key, speak, release to insert), double-tap the same key for hands-free recording, and a toggle shortcut as a keyboard-combo alternative. All ship enabled; configure everything under Preferences → Voice Typing.
The recording pill: red dot + timer and live waveform. In toggle mode it also shows stop and ✕ buttons; in hold mode releasing the key is the stop, so there are no buttons.
Hold-to-talk (push-to-talk)
The fastest way to dictate short bursts: hold, speak, release. No second keypress to remember.
Press and hold your key
Hold the hold-to-talk key on its own for a moment (about a quarter second). The recording pill appears with a timer and a live waveform — recording never happens without it. The microphone is not touched until the hold qualifies, so quick taps and keyboard chords never blink the mic indicator.
Speak — and keep working
Dictate what you want to write. With Fn, Right ⌘ or Right Ctrl you can click and scroll while speaking — useful for placing the cursor exactly where the text should land. Pressing any other key while holding cancels instead (that is a real keyboard shortcut, and your shortcuts always win).
Release to insert
Let go of the key. Colabio transcribes your speech and types it at your cursor. If an app blocks simulated pasting, the text lands on your clipboard and the pill tells you to press paste. Released too early by accident? Nothing happens — a single tap under the threshold is discarded silently (two quick taps start a hands-free session — see below).
Which key, on which OS
Only keys that do nothing on their own qualify — the key still reaches the app you're typing in, so a letter key would both dictate and type. Pick yours in Preferences → Voice Typing → Hold-to-Talk Key — or use Record a key there to assign any key your keyboard can send (F13–F24 make excellent dedicated dictation keys).
macOS
A truly dedicated key — it does nothing on its own, so there is zero conflict with shortcuts. Requires a one-time macOS setting (below). While dictating with Fn you can click and scroll freely; release to insert at wherever your cursor is now.
On every Apple keyboard. While holding it you can click and scroll freely — aim your cursor, release to insert. Keyboard chords stay safe: pressing any other key while holding cancels dictation instantly, so ⌘C behaves exactly as normal.
Windows
Present on effectively every PC keyboard and almost never used in chords (shortcuts are typed with the left Ctrl). While holding it you can click and scroll freely — place the cursor while you speak. The Fn key is not an option on Windows: PC keyboards handle Fn inside the keyboard firmware, so it never reaches the operating system — no application can use it, ours or anyone else's.
Linux (X11)
Same behavior as Windows: hold to record, release to insert, any chord cancels.
Linux (Wayland)
Wayland blocks global key monitoring by design — an industry-wide restriction that affects every dictation and tracking tool, not a Colabio limitation. The setting shows a clear error instead of failing silently. The toggle shortcut may also be unreliable depending on your compositor.
Your shortcuts always keep working
- Pressing the key together with anything else — a key, a click, a scroll — never starts dictation. The microphone is only opened after the key has been held alone past the threshold.
- So ⌘C, Ctrl+C and every other keyboard chord behave exactly as they always have — no pill flash, no mic indicator blink.
- Clicking and scrolling during a dictation is allowed on the default keys (Fn, Right ⌘, Right Ctrl): mouse chords are made with the left-hand modifiers, so the right-side keys stay free for aiming your cursor while you speak. Left-side modifiers keep the click-cancel.
Hands-free: double-tap or the toggle shortcut
For longer dictations where holding a key is tiring — start it and speak as long as you like. Start it either by double-tapping your dictation key (two quick taps of Fn on Mac) or by pressing the toggle shortcut.
Double-tap your key — or press the shortcut
Double-tap your dictation key (two quick taps of Fn on Mac, Right Ctrl on Windows/Linux) — or press ⌘ ⇧ Space on Mac / Ctrl + Shift + Space on Windows/Linux (or your own custom shortcut — see below). The pill appears and recording starts, hands-free: no key to hold.
Stop — or cancel
Tap your dictation key once, press the shortcut again, or click the pill's white stop button, and the transcript is typed at your cursor. The ✕ button (or Esc) cancels and discards the recording. In hands-free mode the pill shows both buttons; in hold mode it shows none — releasing the key is the stop.
Make the shortcut yours
The default (⌘⇧Space / Ctrl+Shift+Space) can collide with other apps' shortcuts. Change it in Preferences → Voice Typing → Dictation Shortcut: click the shortcut chip, press your new combination — any modifier+key combo, or a lone function key — and it's saved instantly. Colabio validates the combo first and tells you if another app already owns it, so you can never save a dead shortcut. “Reset to default” is one click away.
Setup by operating system
The flow is identical everywhere — what differs is the microphone permission, the transcription engine, and (on macOS with Fn) one system setting.
macOS
⌘ ⇧ Space- The first time you dictate, macOS asks for microphone access — click Allow. If you declined it, re-enable it under System Settings → Privacy & Security → Microphone. Hold-to-talk also uses the Accessibility permission Colabio's tracking already has — no new prompt.
- Using the Fn key? One-time setup: macOS assigns the Globe key a system action (emoji picker or input-source switching) by default. Set System Settings → Keyboard → “Press 🌐 key to” → “Do Nothing”. Colabio detects this automatically and shows a reminder in Preferences until it's set.
- Macs get on-device transcription: a small speech model downloads once in the background, and after that your audio is processed locally and never leaves your computer. Until the download finishes, dictation uses the cloud engine automatically. Switch engines any time from the app's ⋮ menu under Voice Typing.
Windows
Ctrl + Shift + Space- If nothing records, check Settings → Privacy & security → Microphone and make sure “Let desktop apps access your microphone” is on.
- Make sure the right input device is selected as your default microphone in Windows sound settings — Colabio records from the system default.
- Transcription runs on Colabio's cloud engine: audio is streamed for transcription only, never stored permanently, and never visible to your organization.
Linux
Ctrl + Shift + Space- Works out of the box with PulseAudio and PipeWire — there's no OS permission prompt; Colabio records from your default input device.
- X11 sessions get the full experience. On Wayland, hold-to-talk is unavailable (the OS blocks global key monitoring for all apps) and the toggle shortcut depends on your compositor.
- Transcription uses the cloud engine, same privacy rules as Windows: transcription only, no permanent storage, invisible to your organization.
On-device vs. cloud transcription
On-device (macOS)
Speech is transcribed locally by a model running on your Mac. Audio never leaves your computer and no outside service is involved — including for Hindi, which is turned into English on your own machine. The model downloads once in the background (dictation uses the cloud until it's ready), then Colabio switches to on-device automatically. Pick it explicitly from the app's ⋮ menu → Voice Typing → On-device.
The trade-off is polish: the cloud engine also tidies your wording, and on-device does not. If your Mac cannot run the local model on a given attempt, Colabio falls back to the cloud for that dictation.
Cloud (all platforms)
Audio is sent to Colabio's transcription service, turned into text, and discarded — never stored permanently, never used for anything but your transcript, and never visible to your organization. This is the default on Windows and Linux, and the automatic fallback on macOS.
The text is then cleaned up before it reaches your cursor: filler words and false starts are dropped, corrections you speak out loud are applied, punctuation is added, and anything said in Hindi comes back in English. Your meaning is kept — nothing is added — but the wording will read tidier than you spoke it. The service uses specialist providers for these steps; see the Privacy page for exactly what each one receives.
Troubleshooting
My hold-to-talk key does nothing
Open Preferences → Voice Typing and use the built-in key test: press your key and watch for the green “Key detected” light. If it never lights up, another dictation or hotkey app (Aqua Voice, superwhisper, Wispr Flow, …) is capturing the key before Colabio can see it — Colabio names the app it recognizes in a red warning. Quit that app or give the two tools different keys. On macOS with Fn, also complete the “Press 🌐 key to: Do Nothing” setting shown above.
Pressing Fn opens the emoji picker
That's the macOS Globe-key system action, which runs alongside dictation until you disable it: System Settings → Keyboard → “Press 🌐 key to” → “Do Nothing”. Colabio shows this exact reminder in Preferences until the setting is correct.
The pill appears but the waveform stays flat
Your microphone isn't delivering audio. Check the OS-level microphone permission (see your platform above) and that the correct input device is the system default, then try again.
Double-tapping my key doesn't start hands-free dictation
The two taps must land within about half a second, and each tap must be short — a long first press reads as hold-to-talk instead. Make sure the key test in Preferences → Voice Typing lights up when you press the key (if not, another app is capturing it), and on macOS set “Press 🌐 key to: Do Nothing” plus turn off the system Dictation shortcut if it's set to “Press 🌐 twice” (System Settings → Keyboard), so Apple's dictation doesn't race Colabio's.
Nothing happens when I press the toggle shortcut
Check that Voice Typing is enabled (Preferences → Voice Typing, or the tray-menu checkbox). If it's enabled, another app may own that combo — set a different one from Preferences → Dictation Shortcut: click the shortcut, press new keys, done. Colabio warns you if the combo is already taken.
My text landed on the clipboard instead of in the app
A few apps (especially ones running with elevated permissions) block simulated pasting. The transcript is already on your clipboard — just press Cmd/Ctrl+V.
Recording stopped on its own
Very long sessions end automatically. Everything you said up to that point is transcribed and inserted as normal — nothing is lost — so just start another session to carry on.
Transcription failed — do I have to say it all again?
No. When a dictation fails for a reason worth retrying — a dropped connection, a hiccup on our side — the pill says “press your shortcut to retry” and keeps the recording for 20 seconds. Press the shortcut once and the same audio is sent again, so you never re-speak a long dictation because of a blip. The audio is held in memory only for those 20 seconds and is discarded as soon as it succeeds, you start a new recording, or the 20 seconds pass — it is never written to disk. If the recording was simply too long, the pill tells you that instead, since retrying could not help.
Hold-to-talk is grayed out on Linux
You're on a Wayland session. Wayland blocks global key monitoring for every application by design. Log into an X11 session to use hold-to-talk, or use the toggle shortcut if your compositor supports it.
Privacy, in one paragraph
Voice Typing records only while you are actively dictating — holding the key, or between your start and stop — and the on-screen pill with its live waveform is visible the entire time: recording is never silent. In hold-to-talk mode the microphone isn't even opened until your hold passes the threshold, so taps and shortcuts never touch it. Audio is used solely to produce your transcript, is never stored permanently, and is completely separate from time tracking: your organization never sees, hears, or receives it. Full details are in our Privacy Policy.