Release Notes

What's new in Voiceavel for macOS.

v1.3.6 Latest

Last Word No Longer Cut Off — Streaming Edition

Real-time streaming with Soniox and AssemblyAI now reliably captures the final word, even when you release the hotkey the instant you finish speaking. Previously, the streaming pipeline closed the WebSocket too eagerly on key release, dropping any audio that hadn't been delivered yet. The connection now stays open for a brief tail period so the trailing audio reaches the recogniser before the session closes — matching the behaviour batch transcription has had since 1.2.

Faster Command Mode

Command mode (select text → hold hotkey → speak instruction) now finalises in about half the time. The session-close wait had quietly grown to 4 seconds across recent builds; we've taken it back down to 2 seconds, matching v1.3.7. Most short commands now complete in roughly a second from when you release the hotkey.

Smarter Command Mode

The AI behind command mode is now llama-3.3-70b-versatile (was llama-3.1-8b-instant). The smaller model was fast but wrote off-script — paraphrasing instead of transforming, echoing the prompt as preamble, dropping content. With 70B in the loop, "make this professional" produces a genuine rewrite, "translate this to Serbian" actually translates, and "fix grammar" leaves your meaning alone. The cost is roughly 200–300 ms per command, well worth it.

Bug Fix: Cursor Check in Browsers

The "cursor not in a text field" check was too lenient on web pages — any focused element that responded to accessibility queries was treated as editable, even when you weren't in an input. The check now requires positive evidence: a non-empty selection, an explicit editable hint, or a focused descendant that's actually a text input. Fewer false positives in Chrome and Safari.

114.8 MB Download
v1.3.5

v1.3.5

  1. Added ability to turn on/off audio support
  2. Made nav icon crispy clear :)
114.8MB
v1.3.4

1.3.4

Launch at Login — Voiceavel can now start automatically when you log in to your Mac. Enable it from Settings or during onboarding.

Fixed UI freeze on connection drops — Stopping a recording while offline no longer freezes the app. Audio and WebSocket shutdown now run in the background with timeouts.

Fixed hotkey leaking to terminals — The Right Command hotkey no longer types a stray "c" character in Ghostty, iTerm, and other terminal apps.

Fixed double word counting — Daily word counter was counting each transcription twice. Now reports accurate usage.

Improved onboarding clarity — The activation screen now clearly states that Voiceavel is free and no credit card is needed.

121.3
v1.3.3

1.3.3

  • Fixed Launch at Login not working — The ServiceManagement framework wasn't being bundled into the built app, causing the feature to silently fail. Now properly included.
121.3 MB
v1.3.2

1.3.2

Launch at Login — Voiceavel can now start automatically when you log in to your Mac. Enable it in Settings, or choose during onboarding. Uses macOS native SMAppService (requires macOS 13+).

Fixed UI freeze on connection drops — Previously, if your internet dropped mid-recording, stopping the recording could freeze the app for up to 20 seconds while waiting for the WebSocket and audio stream to close. The stop action is now fully non-blocking — audio stream shutdown and WebSocket close run in background threads with timeouts, so the UI stays responsive no matter the network state.

Improved stream reliability — The Soniox transcription engine now safely nulls the WebSocket connection before closing the audio stream, preventing the audio callback from sending on a dead socket. Also fixed a Unicode encoding issue in debug logging.

121.3 MB
v1.3.1

**v1.3.1 **

New:

  • Account email now visible in the License menu — easily see which
    account is active on this device
  • Launch at Login — Voiceavel can now start automatically when you log in (Settings → Launch at Login)

Fixes:

  • Fixed app freezing during recording on unstable connections
  • Improved audio stream shutdown reliability
  • Fixed Soniox streaming hang when stopping mid-recording
121.3 MB
v1.3.0

Voiceavel v1.3.0
*Fixed: App freezing during recording *
The app could become completely unresponsive (frozen menu bar icon, no clicks, can't quit) when network conditions degraded during a Soniox streaming session.

What happened: When releasing the hotkey to stop recording, the app waited for the audio stream's callback to finish. If that callback was stuck trying to send audio over a slow/degraded WebSocket, the entire main thread (and UI) would deadlock indefinitely.

**What's fixed: **

  • Audio stream now shuts down immediately without waiting for in-flight callbacks
  • The transcription engine stops accepting new audio before the stream is closed
  • Audio sends bail out instantly when a recording is ending

The menu bar icon will now always return to idle and remain responsive, even on unstable connections.

129.3
v1.2.9

****v1.2.9

Minor garbage colection changes.

121.2 MB
v1.2.8

v1.2.8

Some button missing background fix

121.6 MB
v1.2.7

v1.2.7

Splash windows visibility fix.

121.6 MB
v1.2.6

v1.2.6

Real-time streaming rebuilt from the ground up. Soniox now types words as you speak without mid-word garbage or backspace jitters — clean, instant text appearing at your cursor.

What's new:

  • Soniox streaming — Stable word-boundary emission eliminates "fr esh us er" artifacts
  • Trial experience — 3 days of Soniox real-time (10 min/day), auto-falls back to local Whisper when trial expires
  • Clearer access rules — BYOK providers (OpenAI, AssemblyAI, Soniox own key) now require Pro; free plan = local only
  • First-run polish — License validation happens before showing upgrade prompts; no more trial screen for existing Pro users
121.6 MB
v1.2.5

Voiceavel 1.2.5

Bug fixes and improvements based on user feedback.

• Trial activation now works correctly — Soniox real-time streaming starts as expected after clicking "Start 3-Day Trial" • Switching providers and cancelling the API key dialog no longer breaks transcription — app automatically falls back to your previous working provider • "Stay on Free" now properly resets to local Whisper, no more delays or silent failures • "No credit card required" label added to the trial button • Users who already used their trial will no longer be shown the trial offer again • What's New screen now appears correctly for every new version • Diagnostic log added at ~/.voiceavel/voiceavel.log — use "Send Logs to Support" from the menu if something isn't working****

129.7 MB
v1.2.4

Voiceavel 1.2.4

• Fixed: Trial activation now works correctly for existing users — Soniox real-time streaming will actually start after clicking "Start 3-Day Trial" • Fixed: Switching to a provider that needs an API key and then cancelling no longer breaks transcription — app rolls back to previous working provider automatically • Fixed: "Stay on Free" now resets to local Whisper so there's no delay or silent failure • Added: "No credit card required" label on the trial button • Fixed: Users who already used their trial won't be shown the trial offer again

124 MB
v1.2.3

Voiceavel v1.2.3

  • Fix post-processing replacing text in wrong app when user switches windows after dictating (e.g. YouTube rewind bug)
  • Fix ghost text re-entry when user sends message before Pass 2 correction completes
  • Add terminal/IDE-aware text replacement (backspace instead of Shift+Left Arrow for terminals and JetBrains IDEs)
  • Add PhpStorm to code editor detection
  • Fix Soniox/Groq server-provided API key activation flow
  • Refresh engine credentials automatically after license validation
  • Initialize license manager before transcription engine for proper key loading
  • Increase paste delay for reliability
  • Bump version to 1.2.3
124 MB
v1.2.2

v1.2.2 — Soniox & Cloud Startup Fix

Soniox real-time streaming and Cloud transcription could fail silently after app launch — the transcription engine was initialized before the license manager, so server-provided API keys weren't available yet. This release fixes that ordering and adds automatic retry logic so credentials are always picked up, whether from cache or a fresh server validation.

Changes:

  • Fixed engine initialization order so server-provided keys (Soniox, Cloud signing secret, Groq) are available immediately on startup
  • Background license validation now refreshes the engine automatically when fresh credentials arrive
  • Selecting Soniox or Groq as provider now attempts a server validation first (like Cloud already did) instead of immediately prompting for an API key
124 MB
v1.2.0

Voiceavel v1.2.0

Real-time Streaming — Text now appears as you speak. Soniox is the new primary streaming engine — words show up live in any text field with ~200ms latency. No API key needed. When you release the hotkey, text is cleaned up instantly (~300ms) before AI processing kicks in.

Pro users get 30 minutes of Soniox streaming per day. Free trial users get 10 minutes per day during the 3-day trial. After the trial ends, free users fall back to local Whisper only. When the daily cap is reached, Voiceavel automatically switches to Groq — no interruption.

Want unlimited Soniox? Bring your own key under Settings → API Key and the daily cap is removed entirely. Soniox charges $0.10/hr billed directly to you.

New Providers — Soniox (real-time, ~200ms, English, included with Pro at 30 min/day or unlimited with BYOK), AssemblyAI (real-time, ~300ms, 6 languages, BYOK at $0.15/hr), and Groq (batch, ~0.9s, 100+ languages, included with Pro unlimited).

New Pricing — Pro Monthly at $7.99/month, Pro Annual at $69/year (save 28%), and Pro Lifetime at $199 one-time. All Pro plans include Soniox streaming, Groq cloud, Smart Cleanup, Command Mode, Context Formatting, and unlimited words.

Command Mode + Streaming — Command mode now works seamlessly with real-time streaming. Select text, hold the hotkey, speak your instruction — the selection stays intact while your command is captured, then applied.

Faster Text Replacement — Clipboard paste replaces character-by-character typing (no more garbled output). Clean text replaces on-screen within ~300ms before AI processing. Session finalization dropped from 1.0s to 0.3s.

Bug Fix: Device Reactivation — Fixed a bug where deactivating a device from the dashboard and re-entering the license key would fail. Devices can now be reactivated without issues.

124 MB
v1.1.7

v1.1.7

Fixing small bugs on changing the microhpones.

123.7 MB
v1.1.6

v1.1.6

Sleep/Wake & Audio Reliability

Fixed microphone becoming unavailable after sleep, screen lock, or switching audio devices.

Voiceavel now automatically detects wake events and reinitializes audio — no restart needed. Also fixed intermittent missing or double start/stop beep sounds.

123.7 MB
v1.1.5

v1.1.5

Fixed an issue where AI commands (translate, summarize, make professional, etc.) would silently fail for some Cloud users — the text was pasted back unchanged. The cause was stale credentials in the AI processor. Also fixed the upgrade welcome dialog showing "v1.1.2" instead of the current version, and switching to OpenAI or Groq now immediately opens the API key dialog instead of showing an easy-to-miss notification.

123.7 MB
v1.1.4

v1.1.4 Fixed a bug where the microphone would stop working after rapid-fire recordings, especially when using AI features like translation or command mode. The issue was a race condition where a previous transcription's cleanup could kill an active recording session. Also improved Whisper's language detection accuracy when set to Auto — short phrases are less likely to be transcribed in the wrong language.

123.7 MB
v1.1.3

v1.1.3 Fixed auto-update so it works on the first attempt — previously required two tries due to a macOS quarantine issue. Added better error diagnostics for AI post-processing on the Cloud provider. Reduced Whisper's tendency to transcribe in the wrong language when set to Auto by adding a server-side prompt hint.

123.7 MB
v1.1.2

v1.1.2 — Reliability, Usage Dashboard & AI Fixes

This release focuses on stability and reliability — fixing cold-start issues, accurate cloud usage tracking, and restoring AI post-processing features.


Cloud Usage Dashboard

Real-time cloud quota display right in the menu bar — no need to check the website.

  • Shows minutes used today and this month (e.g. "12m 30s / 60 min")
  • Data comes directly from the server — always accurate
  • Updates automatically after each cloud transcription

Dictation History

Rolling log of your last 30 transcriptions, accessible from the menu bar.

  • Click any entry to copy it to clipboard
  • "Copy Last" shortcut for quick access
  • History persists across app restarts

Audio Feedback

Optional start/stop beep sounds so you know when recording begins and ends.

  • Uses native macOS sounds (Tink/Pop)
  • Toggle on/off under Settings > Audio Feedback

Automatic Update Checks

The app now checks for new versions every 5 minutes while running. No more manually checking for updates.


Bug Fixes

  • Fixed cold-start license failure — First hotkey press after launch no longer shows "Enter API Key" for activated users. The app now validates your license immediately on startup.
  • Fixed cold-start microphone failure — First recording no longer fails with "Microphone not working." The app now waits for CoreAudio to fully initialize before recording.
  • Fixed "No API Key" shown for Cloud provider — Switching to Cloud no longer incorrectly asks for an API key when the signing secret hasn't been cached yet.
  • Fixed cloud minutes showing wrong numbers — Removed local minute tracking that drifted from the server. All usage data now comes exclusively from the server.
  • Fixed double-logging of cloud transcriptions — Cloud transcriptions were being counted twice in usage logs. Now logged once by the server only.
  • Fixed AI features not working on Cloud provider — Translation, Command Mode, Smart Cleanup, and other AI post-processing features were silently failing because the cloud credentials weren't being passed to the AI processor. Fixed.
  • Fixed AI features breaking on provider switch — Switching providers now properly resets the AI processor so it picks up the correct credentials.
123.7 MB
v1.1.1

Built-in Cloud AI Transcription

Cloud transcription is now built into Voiceavel — no separate API keys needed. Powered by Groq Whisper Large v3 Turbo (216x real-time speed), included free for all Pro users.

Pro users can still bring their own OpenAI or Groq API keys under Settings > Provider.

License Activation

Every user now needs a license key. Register free at voiceavel.com/register, get your key by email, and paste it on first launch. All new users get 3 days of full Pro access to try everything out.

Plan Enforcement

  • Free — Local Whisper transcription + AI post-processing, 1 device
  • Pro — Everything in Free + Cloud AI, OpenAI/Groq (own key), up to 3 devices, priority support

Existing Users (v1.0.x / v1.1.0)

Your app will notify you about this update. You'll need to register and enter your license key — your existing settings are preserved. Free users get a 3-day Pro trial automatically.

Bug Fixes

  • Fixed Cmd+V paste not working in the license key field
  • Fixed trial registration overwriting existing license keys
  • Improved onboarding flow and error messaging
123.7 MB
v1.1.0

v1.1.0 — AI-Powered Voice Intelligence

Voiceavel 1.1.0 is our biggest feature update yet. Your voice input now goes through a smart AI post-processing layer that cleans up, corrects, and formats your text automatically — all in under half a second.


Smart Cleanup

Removes filler words and handles mid-sentence corrections automatically.

Filler word removal:

You say You get
"So um I was thinking we should uh meet on Tuesday" "I was thinking we should meet on Tuesday"
"The project is basically, you know, almost done" "The project is almost done"
"I mean, it would be great to sort of get started" "It would be great to get started"

Voice corrections — change your mind mid-sentence:

You say You get
"Let's meet tomorrow, no wait, Friday" "Let's meet Friday"
"Send it to John, actually, send it to Sarah" "Send it to Sarah"
"The deadline is March 10th, scratch that, March 15th" "The deadline is March 15th"

Works with your existing OpenAI or Groq API key. Local-only users still get basic filler removal without needing any API key.


Command Mode

Select any text, hold your hotkey, and speak a command. Voiceavel detects the selection automatically — no extra hotkey needed.

Selected text You say Result
"hey can u send me the doc asap thx" "make this professional" "Hi, could you please send me the document at your earliest convenience? Thank you."
"The quarterly results show a 15% increase in revenue" "translate to Spanish" "Los resultados trimestrales muestran un aumento del 15% en los ingresos"
"We need to implement the new feature by Friday and also fix the bug in the login page and update the documentation" "break into bullet points" "- Implement the new feature by Friday\n- Fix the bug in the login page\n- Update the documentation"
"This is a really really important meeting that we absolutely need to attend" "make it shorter" "This is an important meeting we need to attend"

If nothing is selected → normal dictation. If text is selected → command mode. Fully automatic.


Context-Aware Formatting

Detects which app you're typing in and adjusts output to match.

You say: "hey can we push the standup to 3pm today something came up"

App Output
Slack / WhatsApp hey can we push the standup to 3pm today, something came up
Mail / Outlook Hi, can we push the standup to 3:00 PM today? Something came up.
Word / Pages Can we push the standup to 3:00 PM today? Something has come up.

Supports 30+ apps out of the box including Slack, WhatsApp, Telegram, Discord, iMessage, Mail, Outlook, Spark, Word, Pages, Notion, and Obsidian.


Coding Mode

In code editors, Voiceavel preserves technical terms and formats them appropriately.

You say You get
"define a function called get user profile" "define a function called getUserProfile"
"set max retry count to 5" "set maxRetryCount to 5"
"import react from react" "import React from 'react'"

Recognized editors: VS Code, Cursor, Xcode, IntelliJ, PyCharm, WebStorm, Sublime Text, Warp, iTerm2, Terminal, Alacritty, Kitty, Nova, Windsurf, CodeRunner.


AI Features Menu

All AI features can be toggled individually under Settings → AI Features:

  • Smart Cleanup — filler removal + voice corrections
  • Context-Aware Formatting — app-specific tone and punctuation
  • Coding Mode — technical term preservation in code editors

Under the hood

Uses your existing API key — same one for transcription. Calls gpt-4o-mini (OpenAI) or llama-3.1-8b-instant (Groq). Cost: ~$0.00003 per dictation. Local-only users get regex-based filler removal without any API key.

123.7 MB
v1.0.3

v1.0.3

New features

About section — New "About" submenu in the menu bar shows your current version number and a link to voiceavel.com.

One-click auto-update — Clicking "Update to vX.X.X" now downloads and installs the update automatically in the background — no more manually dragging to /Applications. The app relaunches itself after a successful update.

Smart version detection — Update checks now use proper semantic version comparison, so the app only prompts you when a genuinely newer version is available.

Bug fixes

False update notification — Fixed an issue where the app would show "Update available" for the version you already had installed.

Duplicate update menu items — Old update entries are now properly cleaned up before adding new ones.

Speech clipping on key release — Recording now continues for an extra 400ms after key release to capture the tail end of speech.

123.6 MB
v1.0.2

Voiceavel v1.0.1 — Release Notes

Release date: March 5, 2026

What's fixed

Speech clipping on key release — The most noticeable issue in v1.0.0 was that the last word (or words) of a sentence would get cut off when releasing the hotkey. This happened because recording stopped the exact instant the key was lifted, before the final audio had a chance to be captured. In v1.0.1, Voiceavel now keeps the microphone collecting audio for an extra 400ms after you release the key. This "tail buffer" ensures your full sentence makes it into the transcription, even if you release the key a split second before finishing your thought. This is the same approach used by push-to-talk systems in apps like Discord and TeamSpeak.

What's included (carried over from v1.0.0)

  • Multi-provider transcription: OpenAI (gpt-4o-transcribe, gpt-4o-mini-transcribe, whisper-1), Groq (whisper-large-v3-turbo, whisper-large-v3), and fully offline Local Whisper (tiny through large-v3)
  • Serbian script transliteration: Automatic, Latin, or Cyrillic output for Serbian language transcription
  • Configurable hotkey: Right Command, Left Command, Right Option, Left Option, or Fn
  • 11 supported languages: Auto-detect, English, Serbian, Spanish, German, French, Italian, Portuguese, Russian, Japanese, Chinese
  • License system with free tier (daily word limit), Pro tier (unlimited), offline grace period, and over-the-air update notifications
  • First-run onboarding wizard with guided provider setup and local Whisper auto-installation
  • Audio device hot-swap — automatically detects headphones/mic plugged or unplugged and reconnects the audio stream
  • Always-on audio stream at the mic's native sample rate with automatic resampling to 16kHz for transcription
  • Menu bar app with animated recording indicator and processing status

Build info

  • Bundle version: 1.0.1
  • Minimum macOS: 13.0 (Ventura)
  • Bundle ID: com.codingwisely.voiceavel
123.6 MB
v1.0.1

v1.0.0 — Initial Release

Features

Multi-provider transcription — Choose between OpenAI (gpt-4o-transcribe, gpt-4o-mini-transcribe, whisper-1), Groq (whisper-large-v3-turbo, whisper-large-v3), or fully offline Local Whisper (tiny, base, small, medium, large-v3).

System-wide voice input — Hold your hotkey, speak, release — transcribed text is pasted directly into whatever app you're using.

Configurable hotkey — Choose between Right Command, Left Command, Right Option, Left Option, or Fn.

11 languages supported — Auto-detect, English, Serbian, Spanish, German, French, Italian, Portuguese, Russian, Japanese, and Chinese.

Serbian script transliteration — Automatic, Latin, or Cyrillic output for Serbian language transcription.

License system — Free tier with daily word limit, Pro tier with unlimited transcriptions. Supports offline grace period (7 days).

First-run onboarding wizard — Guided setup for provider selection, API key entry, and local Whisper auto-installation.

Audio device hot-swap — Automatically detects when headphones or microphone are plugged/unplugged and reconnects the audio stream.

Always-on audio stream — Records at the mic's native sample rate with automatic resampling to 16kHz for transcription.

Menu bar app — Lightweight macOS menu bar app with animated recording indicator and processing status.

123.6 MB
v1.0.0

v1.0.0 — Initial Release

Features

Multi-provider transcription — Choose between OpenAI (gpt-4o-transcribe, gpt-4o-mini-transcribe, whisper-1), Groq (whisper-large-v3-turbo, whisper-large-v3), or fully offline Local Whisper (tiny, base, small, medium, large-v3).

System-wide voice input — Hold your hotkey, speak, release — transcribed text is pasted directly into whatever app you're using.

Configurable hotkey — Choose between Right Command, Left Command, Right Option, Left Option, or Fn.

11 languages supported — Auto-detect, English, Serbian, Spanish, German, French, Italian, Portuguese, Russian, Japanese, and Chinese.

Serbian script transliteration — Automatic, Latin, or Cyrillic output for Serbian language transcription.

License system — Free tier with daily word limit, Pro tier with unlimited transcriptions. Supports offline grace period (7 days).

First-run onboarding wizard — Guided setup for provider selection, API key entry, and local Whisper auto-installation.

Audio device hot-swap — Automatically detects when headphones or microphone are plugged/unplugged and reconnects the audio stream.

Always-on audio stream — Records at the mic's native sample rate with automatic resampling to 16kHz for transcription.

Menu bar app — Lightweight macOS menu bar app with animated recording indicator and processing status.

123.6 MB