Everything you need to know about WhisperClick, from first install to fixing common issues.
WhisperClick is a desktop voice-to-text app. Press a hotkey from any application, speak naturally, and your transcribed text is pasted right where your cursor is. No window switching, no copying and pasting. It works in email, Slack, VS Code, Google Docs, terminals, and every other text field on your screen.
Yes. WhisperClick is free for personal and non-commercial use under the CC BY-NC-SA 4.0 license. The app itself costs nothing. If you use cloud transcription (API mode), your API provider may charge a small amount per request, but typical usage runs well under $1/month.
| Platform | Status | Download |
|---|---|---|
| Windows | Fully tested and stable | Setup installer (.exe) or portable (.exe) |
| macOS | Early access | DMG for Apple Silicon (M1/M2/M3/M4) and Intel (2015-2020) |
| Linux | Early access | AppImage |
All downloads are on the GitHub Releases page. The app auto-updates after you install, so you only need to download once.
It depends on which mode you use:
There is no telemetry, no analytics, and no background network activity. See PRIVACY.md for full details.
| Local Mode | Cloud Mode (API) | |
|---|---|---|
| Where processing happens | On your computer | OpenAI or Google servers |
| Internet required | No (after initial model download) | Yes |
| Speed | Depends on your hardware | Typically 1-3 seconds |
| Accuracy | Good (varies by model size) | Excellent (state-of-the-art models) |
| Cost | Free | Pay-per-use via your API key (typically under $1/month) |
| Privacy | Audio never leaves your machine | Audio sent to provider for processing |
| Languages | 50+ (Whisper models) | 50+ (OpenAI), 40+ (Gemini) |
Local mode uses faster-whisper models that run entirely on your CPU. Cloud mode sends audio to OpenAI or Google Gemini for transcription using their latest models.
Yes, in local mode. You need to download a Whisper model once (this requires internet), but after that, all transcription happens offline. Open Settings, switch the mode slider to “Local,” and select a downloaded model. Models range from “tiny” (fast, lower accuracy) to “large-v3” (slower, highest accuracy).
WhisperClick supports 50+ languages for transcription. You can either let it auto-detect the spoken language or pick one manually. Supported languages include English, Spanish, French, German, Italian, Portuguese, Japanese, Korean, Chinese, Hindi, Arabic, and many more.
Translation is also supported: speak in one language and get text in another. Set a source and target language in Settings under “Language & Output.”
Both providers walk you through the process. It takes about 30 seconds.
| OpenAI | Google Gemini | |
|---|---|---|
| Best models | GPT-4o Transcribe, Whisper | Gemini 2.5 Flash, 2.5 Pro |
| Accuracy | Excellent, industry standard | Excellent, rapidly improving |
| Speed | Fast (1-3s typical) | Fast (1-3s typical) |
| Free tier | Pay-as-you-go (no free tier, but very cheap) | Generous free tier available |
| Pricing | ~$0.006/min (Whisper), varies by model | Free tier, then pay-as-you-go |
| Best for | Proven reliability, widest language support | Budget-conscious users, Google ecosystem |
Short answer: If you want a free tier to try things out, start with Gemini. If you want the most battle-tested transcription, go with OpenAI. Both work well.
Typical voice-to-text usage (a few dozen short recordings per day) costs well under $1/month with either provider. OpenAI charges per minute of audio. Gemini offers a free tier that covers light usage. Check each provider’s pricing page for current rates:
Yes. API keys are encrypted at rest using Electron’s safeStorage, which delegates to your operating system’s native credential store (Windows Credential Locker, macOS Keychain, or the Linux secret service). Keys are never stored in plain text. If WhisperClick detects a legacy plaintext key from an older version, it automatically encrypts it on the next save.
Your API key is only ever sent to the provider you selected (OpenAI or Google). It is never transmitted anywhere else.
Open Settings and scroll to the “System” section. You have two options:
Ctrl+Alt+W).The hotkey must include a modifier key (Ctrl, Alt, Shift, or Win) or be an F-key (F7-F12). WhisperClick shows a color-coded indicator: green means safe, amber means it might conflict with other apps, and red means it is blocked because it would override essential system shortcuts (like Ctrl+C).
The default hotkey is Ctrl+Alt+R.
The pill is a small floating capsule that sits at the edge of your screen. When you are not recording, it is a tiny 72x14 pixel dormant capsule. When recording starts, it expands to show live audio bars, a stop button, and a cancel button.
You can:
It always stays in sync with the main window and system tray.
When auto-paste is enabled (the default), WhisperClick remembers which application had focus before you started recording. After transcription finishes, it copies the text to your clipboard and simulates Ctrl+V in that application. The text appears right where your cursor was.
You can toggle auto-paste in Settings under “Output.” If you turn it off, transcriptions are still saved to your history and can be copied manually.
Yes. WhisperClick works with any application that accepts keyboard input and clipboard paste. This includes web browsers, email clients, code editors, terminals, chat apps, word processors, and more. The global hotkey and auto-paste operate at the OS level, so they are not limited to specific apps.
WhisperClick is a transcription tool, not a voice command system. It converts your speech to text and pastes it. It does not execute commands, control your computer, or interact with other apps beyond pasting text. If you say “open my browser,” it will type the words “open my browser.”
Common causes:
WhisperClick validates key format when you enter it. If the format looks correct but transcription still fails, the issue is usually on the provider’s side (billing, quota, or region restrictions).
Check these in order:
Correct device selected: Open WhisperClick Settings and check the Microphone dropdown. Make sure the right input device is selected, not a virtual device or a disconnected headset.
Mic not muted: Check that the microphone is not physically muted (hardware switch on headsets) and that the system volume is not at zero.
Windows SmartScreen may show a warning like “Windows protected your PC” when you run the installer. This happens because the app is not yet code-signed with an Extended Validation (EV) certificate.
To proceed:
This is a one-time step. The app is safe. Code signing is on the roadmap (see ROADMAP.md).
If you prefer not to bypass SmartScreen, you can use the portable version instead of the installer, or build from source (see the README).
macOS may show “WhisperClick can’t be opened because it is from an unidentified developer.” This happens because the app is not yet notarized with Apple.
To proceed:
Alternatively, right-click the app in Finder and select “Open” from the context menu. This bypasses Gatekeeper for that specific app.
Apple notarization is on the roadmap (see ROADMAP.md).
If transcription succeeds (you see text in the history) but it does not paste into your target app:
Focus timing: WhisperClick captures which window had focus before recording. If you clicked somewhere else during recording, the paste target may be wrong. Keep your cursor in the target app before pressing the hotkey.
Auto-paste disabled: Check Settings > Output and make sure auto-paste is turned on.
Target app blocks simulated input: Some apps with elevated privileges (admin consoles, certain security tools) may ignore simulated keystrokes. Try pasting manually with Ctrl+V (or Cmd+V on macOS) after the transcription appears in your history.
Local transcription uses your CPU to run the Whisper model. This is expected during processing and should return to normal once transcription finishes.
To reduce CPU usage:
If CPU stays high even when you are not recording or transcribing, restart the app. The Python sidecar process should be idle between recordings.
WhisperClick checks for updates automatically and downloads them in the background. If updates are not being applied:
github.com and objects.githubusercontent.com.%APPDATA%/Electron/whisperclick/settings.json (or whisperclick-beta for the beta channel).~/Library/Application Support/whisperclick/settings.json.~/.config/whisperclick/settings.json.
The app will recreate default settings on next launch.Sidecar not starting: WhisperClick relies on a Python sidecar process for recording and transcription. If it fails to start, the app will show an error. Try restarting the app. The sidecar auto-restarts up to 3 times with exponential backoff.
Antivirus interference: Some antivirus software blocks the Python sidecar process. Add WhisperClick’s installation directory to your antivirus exclusion list.
WhisperClick or python processes and end them.