Voxa Desktop v0.2.4 — macOS (Apple Silicon)

The first macOS build of Voxa. Voice-to-text: press your hotkey in any text field, speak, and your words are transcribed (via ChatGPT) and pasted where the cursor is.

This is the same 0.2.4 desktop app as Voxa Desktop v0.2.4 (Windows), built for macOS with the platform parity work from #3.

Requirements

Apple Silicon (M1/M2/M3/M4) only. This DMG is arm64; there is no Intel build in this release.

Install

1. Download Voxa-0.2.4-arm64.dmg and drag Voxa to Applications.
2. The app is unsigned and not notarized (no Apple Developer account), so Gatekeeper will say it's from an unidentified developer. To get past it, either:
   • Right-click Voxa in Finder and choose Open, then confirm — or
   • Run xattr -cr /Applications/Voxa.app in Terminal, then launch normally.
3. On first launch, grant the two permissions Voxa asks for:
   • Microphone — required to record at all.
   • Accessibility (System Settings → Privacy & Security → Accessibility) — required for auto-paste. Without it, Voxa still works but falls back to clipboard-only, and you paste with Cmd+V yourself.

Using it

• Start recording: Option+Space. (F9 is macOS's reserved Mission Control / App Exposé shortcut on many keyboards, so it isn't the default here the way it is on Windows.)
• Stop: Escape, or click stop on the recording strip.
• Both hotkeys are rebindable under Shortcuts in the Hub.

What's in this build for macOS

• Automatic paste works. Cmd+V is sent via System Events, gated on Accessibility permission, with a clipboard fallback if that permission isn't granted.
• History labels the app you dictated into — VS Code, Slack, Chrome, and so on.
• The hotkey doesn't steal focus. The recording strip appears without activating Voxa, so the app you're dictating into stays frontmost and the Hub window no longer jumps in front of it.
• Menu-bar icon adapts to light/dark mode instead of rendering as a flat colored block.
• UI text reads as Mac. Hotkey labels show Cmd / Option rather than Ctrl / Alt, and mic-permission and startup copy no longer say "Windows".
• Clearer transcribing state. After you hit stop, the strip shows a distinct transcribing… state rather than just dimming the button.

Verify your download

shasum -a 256 Voxa-0.2.4-arm64.dmg
a4ce6871d4fb17b78017c92b0a69e325f3aa942d7a7716721a9a4c8e319c1d23

Known issues

• Intel Macs are not covered by this release.
• In the Hub's Shortcuts panel, the quick-pick chip highlight doesn't track the currently-set hotkey correctly (the Esc chip loses its highlight, and Cmd-based chips don't light up when picked). Cosmetic only — the hotkeys themselves work.

Voxa Desktop v0.2.4

Voice-to-text desktop app. Press your hotkey in any text field, speak, and your words are transcribed (via ChatGPT) and pasted where the cursor is.

What's new in 0.2.4

No more corrupt recordings. Audio chunks are now written in strict order, fixing an intermittent bug where a recording could be saved corrupt (and then fail transcription with no way to recover it).

"No speech" is no longer an error. A muted/silent mic now shows a calm No speech detected — check your microphone instead of a red failure, and doesn't offer a pointless Retry.

Sticky sidebar + pagination. The left nav stays fixed while History scrolls, and long histories page 50 at a time with Load more.

Your profile, bottom-left. Your signed-in ChatGPT account (name, email, avatar) now shows in the sidebar, with Sign out and Re-authenticate right there — no digging into Settings.

Recorded audio is kept. Past dictations stay playable (no more 24-hour purge). Disk is bounded by a 2 GB cap that clears the oldest clips first. The drawer says "audio no longer available" only when it genuinely is.

Clearer label. "Re-insert" is now Paste into focused app, with a tooltip explaining it.

Install

Download Voxa-Setup-0.2.4.exe and run it (Windows SmartScreen → More info → Run anyway, since the app is unsigned). Quit the running Voxa from the tray first, then install over the top — your history, login, and settings are preserved.

Voxa Desktop v0.2.3

Voice-to-text desktop app. Press your hotkey in any text field, speak, and your words are transcribed (via ChatGPT) and pasted where the cursor is.

What's new in 0.2.3

New Hub interface — a sidebar app: Home, History, Microphone, Shortcuts, Settings (parchment theme).

Dictation history — every dictation is saved, searchable, and grouped by day, with a detail panel and audio playback.

Your audio is never lost — audio is written to disk as you speak, so a crash or a failed transcription no longer loses the recording.

Retry failed dictations — a "Last failed dictations" banner with a Retry button re-runs transcription on the saved audio (great for when the network drops mid-dictation).

Unlimited recording — the cap is now 1–10 minutes or No limit (record until you stop). Settings → Recording strip.

Fixes

• Drawer audio now plays with a working timeline.
• Window min/max/close buttons no longer overlap the UI.
• If auto-paste can't land, you're notified and the transcript is guaranteed on the clipboard (no more silent failures).

Install

Download Voxa-Setup-0.2.3.exe and run it. Windows SmartScreen may warn (the app is unsigned) — choose More info → Run anyway. Keep the default location to update over a previous version; your history, login, and settings are preserved.

Voxa Desktop v0.1.0

First public release of Voxa Desktop — a cross-platform Electron app that records your voice with a global hotkey, transcribes it through your ChatGPT session, and auto-types the result into whatever text field is focused.

What it does

• System-wide hotkey listening (works in any app: VS Code, Word, browser, etc.)
• Small floating overlay shows recording timer
• Transcribes via ChatGPT's backend
• Transcript is auto-typed into the focused text field
• Configurable hotkeys, overlay position, max recording duration
• Runs in the system tray, no window clutter

Defaults

• Start hotkey: `Ctrl+Shift+R`
• Stop hotkey: `Escape`
• Overlay position: top-right

First run

1. Install and launch.
2. Right-click the tray icon → Login to ChatGPT.
3. Complete your ChatGPT login in the browser window that opens (handle 2FA / CAPTCHA there).
4. Once signed in, the app captures your session and you're ready.

Downloads

Windows

• `Voxa Setup 0.1.0.exe` — NSIS installer (recommended). Sets up shortcuts, registers uninstall entry.
• `Voxa-0.1.0-x64.exe` — Portable, x64. Run without installing.
• `Voxa 0.1.0.exe` — Portable. Run without installing.

Linux

• `voxa_0.1.0_amd64.deb` — Debian / Ubuntu package. Install with `sudo dpkg -i voxa_0.1.0_amd64.deb` then `sudo apt -f install` if there are dependency complaints.
• `Voxa-0.1.0.AppImage` — Universal Linux. Has known issues — please read the section below before reporting bugs.

Linux known issues (please read)

The AppImage and .deb work on most desktops, but Linux distros vary enormously. Here's what to expect.

AppImage

• Needs the executable bit: `chmod +x Voxa-0.1.0.AppImage`
• Some distros need `libfuse2`: Ubuntu 22.04+ ships without it. Install with `sudo apt install libfuse2`.
• Wayland sessions: global hotkeys may not work under Wayland because of compositor security. If your hotkey doesn't fire, log out and pick an X11 session at the login screen, or wait for an upstream fix.
• Microphone: the system audio settings must grant the AppImage mic access. On PipeWire / PulseAudio this should be automatic; on sandboxed environments (Flatpak / Snap) it can break — Voxa is not sandboxed (we ship as a raw AppImage), so this is rare but possible.
• System tray: KDE Plasma 5.20+ disabled legacy XEmbed tray by default. Enable Settings → Notifications → System Tray → Show legacy tray icons to see the Voxa icon.

.deb (Debian/Ubuntu)

• Auto-launch on login may not work on distros that use systemd user services for autostart instead of the legacy `.config/autostart`. Manual launch from the menu works.
• Mic permission is usually handled by Pulse/Pipewire and "just works" — if no audio is recorded, check your audio system can see Voxa as a recording client (`pavucontrol` → Recording tab).

Both Linux builds

• No code signing. Your distro / SmartScreen-equivalent may warn. Voxa is open source; you can build from source if you don't trust the binary.

Disclaimer

Voxa uses the ChatGPT web backend (not the official OpenAI Whisper API). This is unofficial use of ChatGPT — it may break at any time if OpenAI changes their endpoints or cookie format, and there is a small risk that unusual cookie usage flags your ChatGPT account. Use at your own risk. This is a personal-use open-source project, not a commercial product.

Source

https://github.com/blackscythe123/voxa

Voxa Android v0.1.0

First public release of Voxa Android — a voice keyboard for Android that records, transcribes via ChatGPT, and inserts the result directly into any text field. Built in native Kotlin + Jetpack Compose.

What ships

Two input methods install side-by-side. The system keyboard picker shows both — switch to whichever fits the moment.

🎤 Voxa Dictate (voice-only IME)

• Big mic button with a live waveform during recording
• Press and hold to talk (or tap-to-toggle if Hold to talk is off in settings)
• Pause / Cancel / Send controls
• Bluetooth headset and wired headset mic routing supported

⌨️ Voxa QWERTY + mic (keyboard IME)

• Full QWERTY layout in Jetpack Compose
• Shift (one-shot + caps lock)
• Symbols layout (`?123`)
• Emoji panel
• Send key that respects `IME_ACTION_*` (Search, Send, Done, etc.)
• Mic button in the top bar — tap to start dictating, tap again to stop and insert the transcript at the caret. You can keep typing on QWERTY while it's recording.

App highlights

• Dashboard home with setup status pill, 2-step onboarding checklist (account / keyboard), and 4 nav tiles (How to use, Settings, Account, About).
• Tutorial screen with step-by-step animated illustrations showing how to enable Voxa in system settings and switch to it from any text field.
• Auth WebView centers the ChatGPT login form vertically and hides marketing chrome (pricing / upsell links) — login itself is never touched.
• Account screen with a working Refresh session button + status pill.
• About with inline privacy text + version info.
• Quiet Sheet visual design (white surfaces, thin strokes, `#1F4FE0` primary, Phosphor-style icons).
• Adaptive launcher icon themed by Android 13+ launchers.

Install (sideload)

The APK is signed with a debug keystore so it sideloads without Play Store. You'll need to allow unknown sources in Android settings.

```bash
adb install app-release.apk
```

Or on the device: download `app-release.apk` → tap to install → grant the "Install unknown apps" permission for your browser.

After install

1. Open Voxa. Grant microphone permission.
2. Sign in to your ChatGPT account in the embedded WebView.
3. Tap Enable keyboard on the home dashboard, then toggle Voxa Dictate and/or Voxa QWERTY + mic on in the system keyboard list.
4. In any text field, long-press the spacebar (or tap the keyboard switcher in the navigation bar) and pick Voxa.
5. Tap the mic, talk, tap again to send.

Requirements

• Android 10 (API 29) or newer
• Internet for transcription (audio uploads to ChatGPT)
• Microphone
• A working ChatGPT account (free tier is fine)
• Optional: Bluetooth permission if you want to dictate through a wireless headset

Known limitations

• The QWERTY keyboard does NOT have word suggestions, autocorrect, or glide / swipe typing. The top-bar suggestion strip is reserved space but stays empty for now. A future release will replace this with a real keyboard engine (HeliBoard fork) to fix this.
• Voice transcription uses ChatGPT's unofficial backend (cookie-based session, same as the desktop app). It may break at any time if OpenAI changes their endpoints, and there is a small theoretical risk that unusual cookie usage flags your ChatGPT account. Use at your own risk.
• No multi-language keyboard layouts yet — US English QWERTY only on the typing IME.
• No clipboard manager, no theme picker, no glide typing yet.

Disclaimer

Voxa is an open-source personal/friends project, not a commercial product. The ChatGPT backend integration is unofficial. We are not affiliated with OpenAI. Use at your own risk.

Source

https://github.com/blackscythe123/voxa