Skip to main content

Your first dictation

This is the one that makes it click. In about five minutes you'll open any text field, hold a conversation with your computer, and watch clean text appear exactly where your cursor was. No copy-paste, no switching windows.

Before you start

This assumes InkSpoke is installed and you've granted microphone (and, on macOS, Accessibility) permission. If not, run through Install and set up InkSpoke first — it takes a couple of minutes.

What you'll need​

  • ~5 minutes and a working microphone.
  • Any app with a text box — a note, an email draft, a chat window, a code editor. A plain text editor is the easiest place to start.
  • Nothing to configure. New installs use the built-in Whisper Small speech model, which runs fully on-device and offline, so your first words never leave your computer.

The whole loop, at a glance​

Press to start, press to finish. That's the entire gesture — everything else is InkSpoke doing the work.

Step by step​

  1. Click into a text field. Put your cursor where you'd normally start typing. This is the spot InkSpoke will type into, so make sure the caret is blinking in a real editable field.

  2. Press the activation hotkey. The default is Alt + Space on Windows and Linux, or ⌥ + Space on macOS. A small listening overlay appears near the center of your screen, and you'll hear a soft start chime.

  3. Speak one sentence. Something natural, like "Thanks for the quick turnaround — I'll review the draft this afternoon." As you talk, a live waveform ripples in the overlay so you can see InkSpoke is hearing you. The label reads Listening… and a timer counts up.

  4. Press the hotkey again — the same Alt + Space / ⌥ + Space. (Prefer the mouse? Click Send in the overlay — it does exactly the same thing.) The label switches to Processing… while InkSpoke transcribes your speech.

  5. Watch it land. The finished text is typed straight into your text field, right where your cursor was. You'll hear a stop chime, and you're back to where you started — ready for the next sentence.

That's it. You just dictated without touching the keyboard.

What the overlay looks like​

The listening overlay is a small, always-on-top window. Here's the anatomy while it's listening:

┌───────────────────────────────────────────────────────────┐
│  ●  Listening…                                    0:03    │
├───────────────────────────────────────────────────────────┤
│   ▁▃▅▇█▇▅▃▂▄▆▇▆▄▂▁▃▅▇▅▃▁▂▄▆▄▂▁                            │
│                                                           │
│   [ Workspace ▾ ]        [ EN ▾ ]              [ Send ]   │
├───────────────────────────────────────────────────────────┤
│   ESC cancel · Alt+Space finish            🎙 Your Mic     │
└───────────────────────────────────────────────────────────┘
  • ● Listening… — the current state. You'll see it move through Preparing…, Listening…, Processing…, and Done (or Error if something goes wrong). If a model needs a moment to warm up the first time, it shows Preparing… briefly before it starts listening.
  • 0:03 — elapsed time for this session.
  • Waveform — a live equalizer that reacts to your voice. Flat bars mean InkSpoke isn't picking up sound; lively bars mean you're coming through.
  • Workspace ▾ and EN ▾ — pickers for the active workspace and dictation language (more on the language picker below).
  • Send — finishes and processes, identical to a second hotkey press.
  • Footer — a reminder of the two keys that matter (ESC cancel · {your finish hotkey}) plus the name of the microphone InkSpoke is using.

What success looks like​

You'll know it worked when:

  • A start chime plays and the overlay appears the instant you press the hotkey.
  • The waveform moves while you speak.
  • On the second press, clean text appears in your field — capitalized and punctuated, not a raw string of lowercase words.

If all three happened, you've got the core loop. Everything else in InkSpoke builds on this.

Now try three things​

You've done the basics. These take seconds each and show off what dictation can do.

1. Toggle AI refinement​

By default, InkSpoke lightly polishes your speech with AI — trimming filler, fixing grammar, and matching the tone of the app you're in. You can flip that on and off without opening settings:

  • Press Alt + Shift + R (Windows/Linux) or ⌥ + ⇧ + R (macOS).
  • When refinement is off, the overlay shows "AI refinement off" and you get the raw, verbatim transcript instead.

Dictate the same rambling sentence twice — once with refinement on, once off — and compare. The difference is the whole point of InkSpoke.

When to turn it off

Verbatim mode is handy when you need your exact words (quoting someone, dictating a password field, or capturing a transcript). For everyday writing, leave refinement on. Learn what it does under the hood in How refinement works.

2. Switch language​

InkSpoke auto-detects your language by default, but you can pin one for a session:

  • With the overlay up, open the language picker (the EN ▾ dropdown) and choose a language, or tap ↑ / ↓ to cycle through it.
  • The list starts with Auto followed by the languages you've added to your preferences.

Pick a language, speak a sentence in it, and finish as usual.

3. Cancel with Esc​

Started dictating and changed your mind? You never have to send it:

  • Press Esc (or click the ✕ on the overlay) to cancel. Nothing is transcribed, nothing is typed, and you'll hear a cancel cue.

Get comfortable with Esc — it's the safety net that makes it easy to just start talking.

If no text appeared​

Almost always, this is the copy-mode fallback doing its job. If, at the moment InkSpoke goes to type, your cursor isn't in a field that can accept text (you clicked the desktop, a button, or a non-editable window), it won't throw your words away. Instead the overlay shows your full transcript with a Copy button and the label "No text field detected."

  • Click Copy, then paste (Ctrl + V / ⌘ + V) where you actually want it.
  • Next time, click into the text field first so the caret is blinking there, then press the hotkey.

A couple of other quick checks if things feel off:

  • No waveform movement? Check that the microphone named in the overlay footer is the one you're speaking into. You can change the input device in Audio & models settings.
  • Overlay never appeared? Another app may have claimed the same shortcut. You can rebind the activation hotkey in General & hotkeys settings.

Power users​

A few things worth knowing once the basics feel natural:

  • Secondary hotkey. There's a second start/stop binding — Ctrl + Shift + Space (Windows/Linux) or ⌃ + ⇧ + Space (macOS) — that behaves identically. Handy if Alt + Space collides with another app.
  • Send equals a second press. The Send button, the activation hotkey, and the secondary hotkey all trigger the same stop-and-process. Use whichever your hands are already on.
  • It's push-to-finish, not push-to-hold. You don't hold the key down while speaking — press once to start, speak freely, press again to finish. See Push-to-talk basics for the full model.

Default hotkeys by platform​

ActionWindows / LinuxmacOS
Start / stop dictationAlt + Space⌥ + Space
Start / stop (secondary)Ctrl + Shift + Space⌃ + ⇧ + Space
Toggle AI refinementAlt + Shift + R⌥ + ⇧ + R
Cancel the overlayEscEsc
Cycle language↑ / ↓↑ / ↓

The activation, secondary, and refinement-toggle hotkeys are all configurable in settings; Esc and the language arrows are fixed.

Next steps​