MacWhisperer Tutorial

📦 Download the App

Download the latest .dmg file from the GitHub Releases page. Open the DMG and drag MacWhisperer.app to your Applications folder.

🛠 Build from Source (optional)

If you prefer building from source, you'll need Xcode 15+ and CMake installed:

On first launch, MacWhisperer opens a setup wizard that walks you through two essential permissions:

🎤 Microphone Permission

MacWhisperer needs access to your microphone to capture audio. macOS will display a system prompt — click Allow. You can verify this later in System Settings > Privacy & Security > Microphone.

♿ Accessibility Permission

This permission allows MacWhisperer to inject transcribed text directly at your cursor position in any application. Without it, the app can still copy text to your clipboard but cannot type it automatically.

1. Open System Settings > Privacy & Security > Accessibility
2. Click the + button and add MacWhisperer
3. Toggle it on

💾 First Model Download

After granting permissions, MacWhisperer downloads the default Whisper model (Small, ~460 MB). This happens once — the model is cached locally for all future use. No network access is needed after this.

MacWhisperer offers five Whisper model sizes. Larger models are more accurate but slower and use more memory. Choose based on your hardware and needs:

To change your model, open Settings from the menu bar icon and select a new model under the Model section. The new model will be downloaded automatically if not already cached.

This is MacWhisperer's core feature: speak, and text appears at your cursor. It works in any application — text editors, email, chat apps, browsers, search fields, and more.

🎹 How to Dictate

1. Place your cursor where you want text to appear (e.g., an email draft, a document, a search field)
2. Hold your keyboard shortcut (configure it in Settings > Insert at Caret). A purple recording indicator appears on screen
3. Speak naturally — don't worry about pace or pauses
4. Release the shortcut — MacWhisperer processes the audio and injects the transcribed text at your cursor position

⌨ Setting the Keyboard Shortcut

Open Settings from the menu bar icon. Under Insert at Caret, click the recorder field and press the key combination you want (e.g., ⌥ Option + Space or F5). This is a hold-to-record shortcut — recording starts when you press and stops when you release.

Streaming mode transcribes audio in real time as you speak, instead of waiting for you to finish. Words appear progressively as they are recognized.

📡 How It Works

When streaming is enabled, MacWhisperer processes audio in small chunks (3-second windows) and injects recognized words immediately. A stability algorithm ensures that only confident words are typed — earlier words may be revised as more context arrives.

⚙ Enable Streaming

Open Settings and toggle Enable streaming mode under the Streaming section. Then use the same hold-to-record shortcut — the only difference is that text appears while you're still speaking.

MacWhisperer can record entire meetings and produce a timestamped transcript with automatic speaker diarization — each participant is identified and labeled.

🎬 Start a Meeting Recording

1. Set up a Meeting shortcut in Settings under the Meeting section (e.g., ⌃ Ctrl + ⇧ Shift + M)
2. Press the shortcut to start recording — the menu bar icon indicates active recording and elapsed time
3. Press the shortcut again to stop recording
4. MacWhisperer processes the audio: transcription + speaker diarization
5. A Markdown transcript is saved to your configured transcript directory

📄 Transcript Output

Meeting transcripts are saved as Markdown files with the following format:

📁 Transcript Storage

By default, transcripts are saved in your Documents folder. Change this in Settings > Transcript location by clicking Choose and selecting your preferred directory.

After a meeting recording, MacWhisperer can optionally send the transcript to an LLM to clean it up: identify speakers by name, fix transcription errors, and improve readability.

⚙ Configure Refinement

1. Open Settings and scroll to the AI Refinement section
2. Toggle Refine transcript with AI on
3. Choose your provider:

✍ Custom Prompts

You can customize the refinement prompt in Settings. The default prompt instructs the LLM to identify speakers by name from context clues, fix misattributed speech segments, and clean up transcription errors. Edit it to match your specific needs — for example, you could add participant names upfront.

Access all settings by clicking the MacWhisperer icon in the menu bar and selecting Settings. The settings panel is organized into four columns:

💡 Best Practices

• Speak clearly at a natural pace — Whisper handles pauses and hesitations well
• Use a good microphone — built-in MacBook mics work, but a dedicated mic improves accuracy significantly
• Match model to task — use Small for daily dictation, Large v3 for important or multilingual recordings
• Quiet environment — background noise reduces accuracy, especially with smaller models
• Set preferred languages — narrowing the language list improves detection speed and accuracy

🔧 Common Issues