VoiceWriter 5.5.0

VoiceWriter 5.5.0

• Dictation runs on Groq (Whisper large-v3) — fast and accurate with your free Groq key. A key saved in the old speak-aloud slot is reused automatically.
• Offline speech with Kokoro 82M (replaces Groq Orpheus and ElevenLabs). Download it once in Settings → Voice Models; it then works with no internet.
• Speech starts in about a second: it plays sentence by sentence as it's generated, and the voice model loads at startup.

Kokoro speaks English only; Bangla text isn't read aloud.

Windows installers updated with the Groq-only dictation build. The Linux packages are from the first 5.5.0 build.

VoiceWriter v0.5.2

• Hide tray icon (Settings → General): run VoiceWriter with nothing on screen. Reopen Settings with voicewriter --settings or by launching the app again.
• Linux packages (.deb, AppImage, .rpm) for Ubuntu 22.04+, Debian 12+, Mint 21+.

VoiceWriter v0.5.1

• Web browser (Alt+W) now reopens exactly where you left off instead of reloading, and remembers your last page across restarts.
• Linux support (Ubuntu 22.04+, Debian 12+, Mint 21+): .deb, AppImage and .rpm packages. On Wayland, bind voicewriter --dictate, --wheel, --speak, --web or --record as desktop shortcuts. Linux packages are attached automatically by CI shortly after this release is published.

VoiceWriter v0.5.0

What's new

• Web browser wedge (Alt+W): a small in-app browser with a custom minimal toolbar (back, forward, address bar, reload, share). Resizable, hidden from the taskbar, and remembers sign-in across restarts.
• AI chat wedge: ask questions with a screenshot, camera photo, or file attached. Model list is fetched live from your Groq account and test-checked automatically. Uses its own API key slot in Settings.
• Call recording: the wheel's Record feature now captures your microphone AND system audio ("what you hear") at once, mixes them into a single saved recording, and produces a chronologically-merged transcript labeled [You] / [System audio]. Falls back to mic-only automatically if system audio isn't available.
• Settings reorganized into collapsible sections (Hotkeys, API Keys & Usage, Transcription Engine, Text Insertion, General), plus a new Guide & API Keys section with step-by-step setup instructions and direct links for every feature.
• Recording confirm window now has fully custom chrome (no native title bar or close button — only Save or Cancel dismiss it).

See the website for the full feature walkthrough.

VoiceWriter v0.3.0

New: Speak selected text aloud

Select text anywhere on your PC, press Ctrl+Alt+C (configurable in Settings),
and VoiceWriter reads it back via Groq text-to-speech. Press the hotkey again to
stop playback mid-clip.

• English only for now.
• Requires its own Groq API key, separate from the dictation key — set it under
  Settings → Speak-aloud API key.
• A live status bar shows progress through capture → request → playback, with a
  real playback percentage once audio starts.

Known limitations

• Groq's current TTS model (Orpheus) is an "expressive" model with no documented
  literal/verbatim mode — very short selections can occasionally produce more
  audio than the text alone would suggest.
• Non-English selections are not blocked; results are unpredictable since
  Orpheus's English voices aren't designed for other languages.

Also in this release

• Application source is now public (previously private) alongside the website.
• Internal: dictation and speak-aloud each keep independent settings.

VoiceWriter v0.2.0

What's new in v0.2.0

Faster — the transcript now appears the instant Whisper returns. The AI grammar/punctuation cleanup runs in the background and silently replaces the text a moment later, so you never wait for it.

Works in more apps — text is now inserted with native Windows input instead of a browser-style method, so it lands in Google Docs / Sheets / Gmail, HubSpot, Aloware and other Chromium/Electron apps that previously ignored it.

Knows code from prose — dictate a shell command and spoken symbols become literals ("dash dash force" -> --force, "slash etc slash hosts" -> /etc/hosts). Commands are kept verbatim, not "corrected". Regular prose still gets full cleanup.

Bangla dictation — hold Alt+X (rebindable) to dictate in Bangla; it types বাংলা script. On Alt+C, romanized "banglish" now passes through untouched instead of being mangled.

More reliable AI cleanup — automatically picks a working Groq chat model, and rejects any cleanup result that mangles or truncates your sentence (keeps the plain transcript instead).

Downloads

• VoiceWriter_0.2.0_x64-setup.exe — NSIS installer (recommended)
• VoiceWriter_0.2.0_x64_en-US.msi — MSI installer

First launch

SmartScreen may warn that the publisher is unknown (not code-signed yet). Click More info -> Run anyway.

Setup

Open the tray icon (near the clock), then paste a free Groq API key from https://console.groq.com for best accuracy, or download a local model from the same panel for fully-offline use.

Windows 10/11, 64-bit.

VoiceWriter v0.1.0

First public release of VoiceWriter.

Hold Alt+C, speak, release — your words are transcribed and typed into whatever field has focus. Runs in the background; open settings from the tray icon.

Downloads

• VoiceWriter_0.1.0_x64-setup.exe — NSIS installer (recommended)
• VoiceWriter_0.1.0_x64_en-US.msi — MSI installer

First launch

Windows SmartScreen may warn that the publisher is unknown (the app is not code-signed yet). Click More info -> Run anyway.

Setup

Open the tray icon (near the clock), then either paste a free Groq API key from https://console.groq.com for best accuracy, or download the local model from the same panel to run fully offline.

Requires Windows 10/11 64-bit.