Skip to main content

Frequently Asked Questions (FAQ)

General​

What is Knowii Voice AI?​

Knowii Voice AI is a privacy-first voice-to-text transcription application that runs entirely on your computer. Your voice never leaves your device by default.

Is it really private?​

Yes! All audio processing and transcription happens locally on your computer by default. No internet connection is required, and your voice data is never sent to any cloud service.

How much does it cost?​

Knowii Voice AI is currently free during the early access period. Future pricing will be announced later.

What operating systems are supported?​

  • Windows: 10 and 11 (fully supported)
  • macOS: 10.13 (High Sierra) or later, on both Apple Silicon and Intel Macs (signed and notarized by Apple)
  • Linux: modern 64-bit distributions via .deb, .rpm, or .AppImage, with Wayland (including Hyprland) and X11 support

Installation & Setup​

How do I download Knowii Voice AI?​

See the Installation Guide for detailed download and installation instructions.

What are the system requirements?​

See the Installation Guide for complete system requirements and hardware recommendations.

How do I update Knowii Voice AI?​

Knowii Voice AI includes automatic updates. The app will notify you when new versions are available. You can also check manually by going to Settings → About → Check for Updates.

For detailed instructions on both automatic and manual updates, see the Updating Guide.

Why do I need to download AI models?​

The AI models are the "brains" that convert your voice to text. They're downloaded separately to keep the initial installer size small and to give you choice over which models to use.

Which model should I choose?​

  • Small models: Faster, use less memory, slightly less accurate. Good for most use cases
  • Large models: Slower, use more memory, more accurate. Best when you need every word right
  • English-only models: Optimized for English-only transcription, typically faster than multi-language equivalents
  • Start with a small model and upgrade if needed

What models are available?​

Knowii Voice AI supports multiple AI transcription models:

Parakeet Models​

  • Parakeet V3 (Recommended): the default choice for most people. Fast and accurate. Supports 25 European languages with automatic language detection. 785 MB download.

Omnilingual Models​

Support for 1,600+ languages with automatic language detection:

  • Omnilingual 300M: Smaller and faster. ~350 MB download
  • Omnilingual 1B: Larger and more accurate. ~985 MB download

Great for multilingual users or less common languages. Still experimental, so accuracy may vary.

Whisper Models (Multi-Language)​

These models support 99 languages:

  • Tiny: Very fast, basic accuracy. 78 MB, ~0.5GB RAM
  • Small: Fast and quite accurate. 488 MB, ~2GB RAM
  • Medium: Accurate but slower. 1520 MB, ~5GB RAM
  • Large V3 Turbo: Accurate but slow. 1620 MB, ~6GB RAM
  • Large V3: Highest accuracy but slowest. 3100 MB, ~10GB RAM

Whisper Models (English-Only)​

Optimized specifically for English transcription:

  • Tiny (English only): Very fast, English-only. 78 MB, ~0.5GB RAM
  • Base (English only): Fast, English-only. 148 MB, ~1GB RAM
  • Small (English only): Fast and accurate, English-only. 488 MB, ~2GB RAM
  • Medium (English only): Accurate, English-only. 1520 MB, ~5GB RAM

Moonshine Models​

Lightweight models designed for computers with limited resources (low memory or CPU). Slower and less accurate than Parakeet or Whisper, but ideal for low-end machines. All Moonshine models use under 1GB RAM:

  • Moonshine Base (English): Better accuracy among Moonshine models. ~239 MB, under 1GB RAM
  • Moonshine Tiny (English): Smallest English model. ~107 MB, under 1GB RAM
  • Moonshine Tiny (Arabic/Chinese/Japanese/Korean/Ukrainian/Vietnamese): Single-language variants. ~143 MB each, under 1GB RAM

Notes:

  • Larger models require more powerful hardware
  • English-only models are optimized for English and cannot transcribe other languages
  • Moonshine models are best for computers with very limited RAM (1GB minimum)
  • Omnilingual models support the most languages but are still experimental
  • You can download and switch between models at any time

Usage​

How do I transcribe my voice?​

See the Getting Started Tutorial for a complete step-by-step walkthrough of your first recording.

Can I change the keyboard shortcut?​

Yes! Go to Settings > General and click on the shortcut to change it to your preferred key combination.

Does it work in all applications?​

Yes! Knowii Voice AI pastes transcribed text into any application that accepts text input: browsers, email clients, word processors, IDEs, chat applications, and more.

How accurate is the transcription?​

Accuracy depends on:

  • Audio quality: Clear audio with minimal background noise works best
  • Speaking style: Clear, natural speech at a moderate pace
  • Model choice: Larger models are generally more accurate
  • Language: Some languages have better model support than others

Typical accuracy is 90-95% for clear audio with good microphones.

Why do numbers appear as words instead of digits?​

Saying "1122" can come out as "one one two two" depending on which AI model you use. The model itself decides how to write numbers. To get digits:

  • Enable "Write digit sequences as numbers" in Settings > Transcription. It converts runs of 3+ spoken digits into numbers, and never touches normal sentences
  • Use Parakeet V2 for English dictation. It writes numbers, dates, and amounts as digits much more consistently than Parakeet V3
  • Speak numbers fluently in a steady rhythm. Long pauses between digits make the model spell them out as words
  • Group numbers naturally when possible (say "eleven twenty-two" rather than "one one two two")

See Numbers and Dates for more tips.

Can I transcribe in languages other than English?​

Yes! Knowii Voice AI supports multiple languages:

  • Omnilingual models: 1,600+ languages with automatic detection
  • Whisper models: 99 languages
  • Parakeet V3: 25 European languages (Parakeet V2 is English only)
  • Moonshine models: English, Arabic, Chinese, Japanese, Korean, Ukrainian, Vietnamese

The default Auto setting automatically detects the language you're speaking and is recommended for most users. Only change the language in Settings > Transcription > Language if you're experiencing issues with automatic detection.

What if the transcription has errors?​

You can:

  • Edit transcriptions directly in the History tab
  • Add Custom Words in Settings > Transcription to help the model with specialized vocabulary
  • Set up Word Replacements for terms that always come out the same wrong way (an exact, guaranteed fix)
  • Set up Phonetic Replacements when a name or acronym comes out in the wrong alphabet
  • Turn on "Remove filler words" to automatically clean up "uh", "um", and stutters
  • Try a more accurate (larger) model

See Transcription Settings for details on each of these options.

Why did my recording produce no text?​

If you tap your shortcut by accident or record without speaking, Knowii Voice AI produces no text on purpose. Very short or silent recordings are ignored rather than transcribed, which prevents the AI model from inventing phrases (like "thanks for watching") out of silence. Just record again and speak normally.

Privacy & Data​

Where is my data stored?​

All data (transcriptions, audio recordings, settings) is stored locally on your computer:

  • Windows: %APPDATA%\knowii-voice-ai
  • macOS: ~/Library/Application Support/knowii-voice-ai
  • Linux: ~/.local/share/knowii-voice-ai (or $XDG_DATA_HOME/knowii-voice-ai if set)

You can open this folder at any time from Settings → About → App Data Folder → Open Folder. See the Application Data page for what's inside, and the Privacy Policy for more details.

Can I delete my history?​

Yes! In the History tab, you can:

  • Delete individual transcriptions
  • Clear all history at once
  • Delete audio files while keeping transcriptions to save space

Does Knowii Voice AI collect telemetry?​

No. Knowii Voice AI does not collect any telemetry, analytics, or usage data. See the Privacy Policy for more details.

Is my audio recorded?​

Audio is temporarily recorded during transcription and optionally saved in your history for playback. You can delete audio files at any time while keeping the transcription text.

Performance​

Why is transcription slow?​

Transcription speed depends on:

  • CPU/GPU: Faster processors = faster transcription
  • Model size: Larger models take longer to process
  • Audio length: Longer recordings take longer to transcribe

To improve speed:

  • Use a smaller model
  • Enable GPU acceleration if you have a Vulkan-compatible GPU (most modern GPUs)
  • Keep recordings short (under 30 seconds for best experience)

Does it use my GPU?​

Yes! If you have a Vulkan-compatible GPU (most modern NVIDIA, AMD, and Intel GPUs), Knowii Voice AI automatically detects and uses it for significantly faster transcription. CUDA support for NVIDIA GPUs is planned for a future release.

Why does it use so much memory?​

AI models require significant memory to run. To reduce memory usage:

  • Use a smaller model
  • Enable auto-unload in Settings > Models to unload models after inactivity

Can I run it on a laptop?​

Yes! Knowii Voice AI works well on laptops. Consider using:

  • Smaller models for better battery life
  • Auto-unload to free memory when not in use
  • On-demand microphone mode instead of always-on

Troubleshooting​

The app doesn't start​

  1. Check system requirements
  2. Verify the application has permission to run
  3. Check system logs for error messages
  4. Try reinstalling the application

No audio is recorded​

  1. Check that your microphone is connected and working
  2. Verify microphone permissions in system settings
  3. Test the microphone in another application
  4. Select the correct microphone in Settings > Audio

The app can't detect my microphone (or the Test microphone button does nothing)​

This almost always means your operating system is blocking apps from using the microphone.

On Windows (the most common case), turn on microphone access for desktop apps:

  1. Press the Windows key, type Microphone privacy settings, and open it.
  2. Turn on Microphone access and Let apps access your microphone.
  3. Scroll down and turn on Let desktop apps access your microphone. This is the key setting for Knowii Voice AI.
  4. Fully close and reopen Knowii Voice AI.

On macOS, allow the microphone in System Settings > Privacy & Security > Microphone and make sure Knowii Voice AI is turned on.

For the full step-by-step guide (including confirming your microphone in your system's sound settings), see Microphone Not Detected.

Good to know: When the microphone is blocked or missing, Knowii Voice AI now shows a notification with guidance on how to fix it, so you're not left guessing.

Shortcut doesn't work​

  1. Check that the shortcut isn't already used by another application
  2. Try a different shortcut in Settings > General
  3. Restart the application after changing shortcuts
  4. On Linux: the shortcut needs keyboard access. Add yourself to the input group with sudo usermod -aG input $USER, then log out and back in. See the Installation guide.

Dictated text doesn't appear (Linux)​

If your shortcut records and transcribes but the text isn't typed into your app, it depends on your desktop:

  • GNOME (Wayland): text output is built in. No ydotool or extra tool to install. If you're on the .deb or .rpm, it works automatically. If you're on the AppImage, run the one-time /dev/uinput command and log out/in once (see the Installation guide link below).
  • KDE, Hyprland/Sway, X11: install the typing tool for your desktop (package names are the same across distributions, so use dnf/apt/pacman/zypper):
DesktopTool
KDE Plasma (Wayland)kwtype
Hyprland / Sway (wlroots)wtype
X11 (any desktop)xdotool

Then restart the app (or pick the tool under Settings → Advanced → Paste → Typing Tool). Knowii Voice AI shows a warning with the exact command when a compatible typing tool is missing. More detail in the Installation guide.

My mouse or keyboard behaves strangely while the app is running (Linux)​

On Linux, Knowii Voice AI listens to your keyboard directly so your shortcut works in every app, on every desktop. While it's listening it takes temporary exclusive control of the devices it recognises as keyboards, so your shortcut doesn't leak into whatever you're typing in.

On rare hardware this can affect the wrong device. Symptoms look like:

  • The mouse pointer stops moving, but clicks still work.
  • A key, media button, or scroll wheel stops responding.
  • Trackpad gestures stop working.

This is most likely with gaming mice and fancy keyboards, which often present themselves as a keyboard and a mouse through the same connection.

To get your device working again right now, quit Knowii Voice AI completely. Control is handed back the moment the app exits. Quit it from the tray icon, or from a terminal:

pkill knowii-voice-ai

You do not need to restart your computer.

To avoid it happening again, first update to the latest version. Devices that also report pointer movement are left alone automatically. If you're already up to date and still hit this, please report it and include:

  • the make and model of the mouse or keyboard affected, and
  • the knowii-voice-ai.log file (Settings → Advanced → Logs), which records exactly which devices the app took control of.
info

This only applies to Linux. On Windows and macOS the app uses the system's own shortcut support and never takes control of your devices.

AppImage: global shortcuts or typing don't work​

On Linux, the AppImage is a single portable file that runs without an installer. Because nothing gets installed, the AppImage can't set up the small system permission rules that Knowii Voice AI needs to (1) detect your global keyboard shortcut and (2) type your transcriptions without a permission pop-up every time. So on the AppImage, your shortcut or the typing may not work until those rules are in place.

The regular .deb and .rpm packages install these rules for you automatically, which is why they "just work".

Easiest fix: install the .deb or .rpm package​

Instead of the AppImage, download and install the package for your system:

  • .deb for Debian, Ubuntu, Linux Mint, and Pop!_OS
  • .rpm for Fedora, RHEL, and openSUSE

These set everything up automatically, so global shortcuts and typing work right after you log in. Get them from the Installation guide or the GitHub Releases downloads page.

Manual fix: if you must use the AppImage​

If you need to stay on the AppImage, you can add the same rules yourself. Open a terminal and paste the following block. It creates the two rules, makes sure the required system module is loaded, and applies everything to your current session:

# 1) Keyboard access for global shortcuts
sudo tee /etc/udev/rules.d/72-knowii-voice-ai-input.rules >/dev/null <<'EOF'
KERNEL=="event*", SUBSYSTEM=="input", ENV{ID_INPUT_KEYBOARD}=="1", TAG+="uaccess"
EOF

# 2) uinput access for prompt-free typing
sudo tee /etc/udev/rules.d/72-knowii-voice-ai-uinput.rules >/dev/null <<'EOF'
KERNEL=="uinput", SUBSYSTEM=="misc", OPTIONS+="static_node=uinput", TAG+="uaccess"
EOF

# 3) Make sure the uinput module is loaded now and on every boot
echo uinput | sudo tee /etc/modules-load.d/knowii-voice-ai-uinput.conf >/dev/null
sudo modprobe uinput

# 4) Apply the rules to your current session
sudo udevadm control --reload-rules
sudo udevadm trigger --subsystem-match=input --action=change
sudo udevadm trigger --subsystem-match=misc --action=change

Then log out and back in (or reboot) and restart Knowii Voice AI.

note

This only affects the AppImage on Linux. The .deb and .rpm packages, as well as Windows and macOS, need none of this. Everything is handled for you.

tip

Once your shortcut works, the in-app warning about it disappears on its own.

System tray icon missing on GNOME​

If you're using the GNOME desktop, you may notice that Knowii Voice AI has no icon in the system tray (usually near your clock). This is not a bug in the app.

Unlike KDE and most other Linux desktops, GNOME does not show tray icons on its own; it removed that built-in feature years ago. To get tray icons back, GNOME users add a small, widely-used add-on called an extension. Once it's installed, the Knowii Voice AI tray icon appears just like it does everywhere else, and the in-app warnings about the missing tray go away.

You only need to do this once.

Option 1: Use the Extensions app (easiest)​

  1. Open the Extensions app (on some systems it's called Extension Manager) from your applications list.
    • Don't have it? Install it first. It's usually called gnome-shell-extension-manager in your software store or package manager. On Ubuntu, the needed support is often already there and just needs turning on.
  2. In the app, search for "AppIndicator and KStatusNotifierItem Support".
  3. Turn the switch next to it on.
  4. The tray icon appears within a few seconds. No restart needed.

Option 2: Enable it from the GNOME Extensions website​

  1. Open this extension page in your web browser.
  2. Flip the toggle at the top of the page to on.
    • The website needs a small browser helper to work. If the toggle doesn't respond, install the GNOME browser integration add-on for your browser. On Fedora, you may also need the gnome-browser-connector package. If this feels fiddly, use Option 1 above instead; it's simpler.
  3. The tray icon appears within a few seconds. No restart needed.

Already installed it and the icon still doesn't appear?​

If the extension is installed and switched on but Knowii Voice AI still says there's no system tray, GNOME is almost certainly blocking it in one of these ways:

  1. All extensions are switched off at once. GNOME has a single master switch that disables every extension, no matter how each one is set. It's silent, and it survives restarts. To check, open a terminal and run:

    gsettings get org.gnome.shell disable-user-extensions

    If it prints true, turn extensions back on with:

    gsettings set org.gnome.shell disable-user-extensions false

  2. A GNOME upgrade disabled it. When GNOME updates to a new major version, it switches off extensions that haven't been marked compatible yet. Open the Extensions app and look for an update, or re-enable it there.

Once extensions are running again, the tray icon appears on its own.

tip

Knowii Voice AI watches for the tray in the background, so whichever option you choose, the icon appears on its own once the extension is running, so you don't need to restart the app. The in-app warning disappears at the same time.

info

The app works fine without the tray icon; you just lose the quick tray menu. Because there's no tray to bring the window back from, Knowii Voice AI automatically protects you so you never lose access to the window:

  • Closing the window minimizes it instead of hiding it. You can bring it back any time from the Activities overview or with Alt+Tab.
  • The Start Hidden option is ignored, so the window always opens when you launch the app.
  • A Quit button appears in the bottom-right corner of the window, since there's no tray menu to quit from.
note

This only affects GNOME when the extension isn't installed or isn't running. KDE and other desktops with a working system tray are not affected, and the tray icon appears normally there. No action needed.

Transcription is completely wrong​

  1. Check the selected language in settings
  2. Try a different model
  3. Test with clear audio and minimal background noise
  4. Verify your microphone is working properly

The app seemed to get stuck while transcribing​

If a transcription takes far too long, the app recovers on its own and returns to a ready state with a clear message. It will not freeze. If you see this repeatedly, try a smaller model or check the Application Logs for details.

My microphone was unplugged (or switched) mid-recording​

Unplugging a USB mic or switching a Bluetooth headset mid-recording no longer breaks anything. The app recovers on its own instead of failing silently. Just reconnect or select a working microphone in Settings > Audio, then record again.

I get an error message when pasting, loading a model, or using the microphone​

Every one of those failures raises a notification that says what happened: text couldn't be pasted, a model failed to load, or the microphone couldn't be accessed. Nothing fails quietly. For more detail, open the Application Logs under Settings > Advanced.

Models won't download​

  1. Check your internet connection (required for downloads only)
  2. Check available disk space
  3. Close and restart the application, then retry the download
  4. Try downloading again (downloads can be resumed)
  5. Try a different model
  6. If the issue persists, Hugging Face (where most models are downloaded from) may be experiencing downtime. Check their status page or try again later

Advanced​

Can I use my own AI models?​

Not currently. Custom model support may be added in a future release. See the Roadmap for planned features.

Can I use this for professional transcription?​

Yes, but please note:

  • Review transcriptions for accuracy
  • Knowii Voice AI is provided as-is without warranties
  • For critical professional use (legal, medical), consider additional review processes

Is there an API?​

Not a programmatic API yet; that stays on the Roadmap. But most automation needs are already covered today: the transcribe CLI scripts file transcription (subtitles, text, JSON, whole folders), and the knowii-voice-ai executable can start, stop, and cancel dictation from a script, a panel button, or a keybinding (Controlling the desktop app).

Is there a CLI?​

Yes. Knowii Voice AI ships with transcribe, a standalone command-line tool that runs fully headless. It transcribes audio and video files to subtitles (srt, vtt) or text, batch-processes entire folders, and manages your transcription models from the terminal. It uses the same engines and the same models as the desktop app, offline, on your machine.

transcribe file talk.mp4 --model whisper-large-v3

That writes talk.srt next to the video.

The knowii-voice-ai executable also takes flags of its own, so you can start or cancel a recording from a script, a panel button, or a window manager keybinding. Those are documented under Controlling the desktop app.

Can I transcribe audio and video files I already have?​

Yes, two ways.

In the app: drag and drop files onto the Knowii Voice AI window, or pick them via Transcribe File... in the tray menu. The transcriptions are saved to your history. See File Transcription.

From the terminal: the transcribe CLI writes subtitles or plain text next to your files, and handles whole folders in one command.

# One video
transcribe file talk.mp4 --model whisper-large-v3

# A whole folder, as text
transcribe file *.mp4 --model whisper-large-v3 --format txt --output ./transcripts/

Both read mp4, mkv, mov, m4a, mp3, wav, ogg, opus, flac, and more, with no ffmpeg needed. See the Transcribe CLI Tutorial to get going with the CLI.

I installed with the MSI. What happens with updates?​

Starting with 0.8.0, Windows releases ship as the setup.exe installer only. The MSI is on hold because of an upstream packaging bug (tauri#14681) that conflicts with bundling the transcribe CLI, and the CLI won that trade.

If you originally installed with the MSI, your next update installs through setup.exe. Everything keeps working; the one cosmetic leftover is the old MSI entry under Windows Apps. Uninstall that old entry once and you are done. The MSI will return when the upstream bug is fixed.

What is the relationship to Handy.computer?​

Knowii Voice AI is a fork of Handy. Handy is licensed under the MIT license, which explicitly allows anyone to use, modify, and redistribute the code. That's exactly what I've done here.

Just because it's a fork doesn't mean it's the same thing. I've recreated the entire frontend, implemented different features, and taken different paths forward. The projects share common origins but have distinct goals and functionality. You can see what's been implemented in the Release Notes and what's planned in the Roadmap.

The decision to keep Knowii Voice AI closed-source comes down to sustainability and the resources needed for ongoing development and support. I intend to contribute improvements back to the original Handy project over time.

This is how open source is supposed to work. Permissive licenses let people build on existing work and take it in new directions.

Still Need Help?​

Can't find the answer you're looking for? Visit the Support page for all available help options including:

  • Email support at support@knowii.net
  • Knowii Community forums
  • GitHub issues and discussions
  • Additional resources and documentation

Tip: When reporting a problem, you can copy your recent logs straight from Settings → Advanced → Application Logs and paste them into your message. See Copy Logs From Inside the App.