Voice dictation & AI writing toolkit for Windows

You won't have to write a single word again.

Click into any field. Hold AltC, say your sentence, let go — it's typed where your cursor was. Select text and press ShiftAltC for a wheel that refines, translates, records a call, opens a mini browser, or starts an AI chat. No window, ever.

Download for Windows See the refine wheel

Free · Windows 10 / 11, 64-bit · v5.5.0

Hold, speak, release.

That's the core interaction. Everything else is one more keypress away.

  1. 1

    Install it. VoiceWriter starts with your PC and sits in the tray — no window on screen.

  2. 2

    Click into any text field, then hold AltC and speak.

  3. 3

    Release. Your sentence appears instantly, then quietly cleans itself up a moment later.

Fast, then polished

The transcript lands the instant it's ready. An AI grammar-and-punctuation pass runs in the background and silently replaces it when done — no waiting.

Works everywhere

Types into stubborn web apps too — Google Docs, Sheets, Gmail, HubSpot, Aloware — using native Windows input, not a browser hack.

Knows code from prose

Dictate a Linux command and "dash dash force" becomes --force; it's kept verbatim, not "corrected". Prose still gets full cleanup.

Two dictation keys

Hold AltC for English (romanized Banglish passes through untouched). Hold AltX for Bangla — types বাংলা script.

Fast, accurate dictation

Dictation runs on Groq's Whisper large-v3 — typically under a second for a sentence, with your own free Groq key.

Invisible & light

Nothing on screen unless you open settings from the tray. Click away and it hides. ~25 MB idle. Keys stored in Windows Credential Manager.

New

Select text. Press one hotkey. Get a wheel of actions.

Select any text on screen and press ShiftAltC — a small radial menu appears right at your cursor. No selection? It still opens; you just get a heads-up if you pick a text action with nothing selected.

Refine

Fixes grammar, spelling, and punctuation only — same meaning, same tone, same length. Won't rewrite what's already correct.

Translate

One-click Bangla, or open the Translate submenu for English, Bangla, Spanish, or Italian. Powered by Groq's chat models with a faithfulness check before anything is shown.

Copied, not pasted

Results are copied to your clipboard and shown in a clean preview panel — nothing is auto-typed into the app you were using. You paste it when you're ready.

New in 5.5

Offline speech, on your PC.

Read-aloud now runs locally with Kokoro 82M. It isn't bundled with the installer — download it once from Settings → Voice Models and it works fully offline after that.

Kokoro 82M speech

Speak-selection and English translations are voiced by Kokoro 82M, a small, natural-sounding open model that runs on your CPU. Pick from several voices and set the speed. One download, about 190 MB.

Starts speaking fast

Speech plays sentence by sentence as it's generated, so you hear the first words in about a second instead of waiting for the whole selection.

Manage everything in one place

Voice Models shows each model's size, download progress, installed status and total storage used. Delete any model in one click to free the space.

Bangla, Spanish & Italian, text-only

These translations appear instantly in the preview panel and copy to your clipboard. Kokoro speaks English, so only English translations are read aloud.

New

Record a call, get a labeled transcript.

Pick the Record wedge from the wheel to capture both sides of a call at once — your microphone and your system audio (what's playing, like the other person's voice). A small, mostly-transparent floating dot follows your cursor while it records, and you can drag it out of the way without stopping the recording.

  1. 1

    Press ShiftAltC and choose Record. A tiny floating icon appears and blinks red while it's listening — drag it anywhere on screen if it's in your way.

  2. 2

    Click it again to stop. A clean confirm window (no title bar, no stray close button — Cancel or Transcribe & Save are the only ways out) shows an audio preview and lets you pick the filename and folder.

  3. 3

    Confirm, and both tracks are transcribed in English, best-effort merged in chronological order and labeled [You] / [System audio], then opened in a dedicated chat-style transcript window. The combined audio and the transcript are both saved to disk too.

One saved recording, not two files

Your mic and system audio are mixed into a single .wav after transcription — you get one normal-sounding recording of the call, not a pair of separate tracks to manage.

System audio is optional

If nothing was playing, or your setup doesn't support capturing it, VoiceWriter falls back to a mic-only recording automatically — it never blocks or fails the whole feature over the optional track.

No Notepad, no clutter

Everything opens in VoiceWriter's own transcript viewer. The .wav and .txt files are still saved to your chosen folder — nothing else pops up.

English-only transcription

Recordings are transcribed in English specifically, rather than auto-detected, so accents and background noise don't cause language mix-ups.

New

A tiny browser, always one hotkey away.

Press AltW to pop open a small in-app browser — no need to alt-tab out to a full browser window just to check something. Press it again to hide the window, press it once more to pick up right where you left off.

Minimal, custom chrome

Back, forward, an address bar, and a share button that copies the current URL — no browser clutter, no native title bar, no stray buttons.

Signed in stays signed in

Email/password sign-in persists across restarts, so you don't have to log back into the same site every time. "Continue with Google" won't work here — that's a Google anti-phishing policy that blocks Google sign-in inside any embedded browser window, in every app that does this, not a VoiceWriter limitation.

Resizable, hidden from the taskbar

Drag any edge to resize it. It never shows up as its own taskbar entry — it's meant to feel like a small utility panel, not another app window.

New

Ask AI, with a screenshot if you need one.

Pick the AI wedge from the wheel for a chat window that can see what you attach — a screenshot, a photo from your camera, or a file — and answer questions about it, powered by your choice of Groq's available chat and vision models.

Pick your own model

The model list is fetched live from your Groq account, not hardcoded — each one is automatically test-checked so you can see which are actually working before you pick.

A separate key, if you want one

AI chat uses its own Groq API key slot in Settings, with its own usage meter — use the same key as dictation, or a completely different Groq account.

Attach and ask

Grab a screenshot, snap a camera photo, or attach a file, then ask a question about it right in the chat — replies render with basic formatting (bold, lists) for readability.

Getting your API keys

Online features run on your own free Groq API key — nothing is billed through the app, and Groq has a free tier with no credit card required. Speech and local dictation need no key at all. Settings also has its own built-in step-by-step guide.

Groq — fast dictation, refine, translate, AI chat, call recording

Used for dictation (Whisper large-v3), the refine/translate wheel actions, the AI chat wedge, and call-recording transcription. Dictation and AI chat each have their own key slot in Settings — reuse one key, or use separate Groq accounts.

  1. 1

    Create a free account at console.groq.com.

  2. 2

    Open console.groq.com/keys and click Create API Key.

  3. 3

    Copy the key (starts with gsk_) and paste it into the matching key field in VoiceWriter's Settings.

Per-key usage meters

Settings shows a live usage bar for Groq dictation and AI chat, tracked against Groq's published free-tier limits, so you always know how much headroom is left.

Download

VoiceWriter for Windows & Linux

Pick your platform below. The offline speech voice is downloaded later from inside the app.

Installer · .exe MSI · for managed installs

Linux (Ubuntu 22.04+, Debian 12+, Mint 21+): .deb · AppImage · .rpm

All releases · Source & issues

Set up your API keys — open the tray (▲ near the clock), click the VoiceWriter icon to open Settings, then:

  1. 1

    Groq (for cloud dictation, refine, translate, AI chat): create a free key at console.groq.com/keys and paste it into the Dictation key field. No credit card needed.

  2. 2

    Voice Models (offline speech): open Settings → Voice Models and download Kokoro 82M for speak-aloud.

  3. 3

    Full walkthrough: API keys section below.

Installing on Linux — Ubuntu 22.04 or newer, Debian 12+, Linux Mint 21+, Fedora and other recent distros (64-bit). Ubuntu 20.04 and older aren't supported.

  1. 1

    Ubuntu / Debian / Mint: download the .deb, then in a terminal: sudo apt install ./VoiceWriter_5.5.0_amd64.deb

  2. 2

    Fedora / openSUSE: download the .rpm, then: sudo dnf install ./VoiceWriter-5.5.0-1.x86_64.rpm

  3. 3

    Any distro, no install: download the AppImage, make it executable with chmod +x VoiceWriter_5.5.0_amd64.AppImage, then run ./VoiceWriter_5.5.0_amd64.AppImage

  4. 4

    Start it: search “VoiceWriter” in your app launcher, or run voicewriter & in a terminal. It runs in the background with an icon in the top bar — click it and choose Open settings to add your API keys. On plain GNOME (Fedora, Debian), install the “AppIndicator and KStatusNotifierItem Support” extension to see the icon.

  5. 5

    Hotkeys on X11 work exactly like on Windows. On Wayland (the default on recent Ubuntu — check with echo $XDG_SESSION_TYPE), apps can't register global hotkeys, so add your own in Settings → Keyboard → Custom Shortcuts using the commands below. For typing text into Wayland apps, also install wtype or ydotool.

Terminal commands (Linux & Windows) — each one controls the running copy, or starts VoiceWriter if it isn't running:

  1. ›

    voicewriter --settings open Settings · --wheel refine wheel · --dictate start dictation, run again to stop and insert · --speak read selection aloud · --web mini browser · --record start/stop call recording

  2. ›

    Hide tray icon (Settings → General) runs VoiceWriter with nothing on screen. Use voicewriter --settings or launch it again to get back to Settings.

Troubleshooting

Common setup issues and how to fix them.

Dictation says “Groq API key required”

Dictation runs on Groq. Create a free key at console.groq.com/keys, paste it into the Dictation key field in Settings → API Keys & Usage, and press Test.

Speak-aloud says nothing

Download Kokoro 82M in Settings → Voice Models, then press Preview to test the voice. Kokoro speaks English only — Bangla text is not read aloud.

Upgrading from 0.5.x

ElevenLabs and Groq speech were removed. Your Groq key still works for dictation and AI chat (a key saved in the old speak-aloud slot is reused automatically); speech now comes from the free offline Kokoro model.