Installation Guide

Everything you need to get Speakeasy up and running on your Mac.

Contents

Quick Start

Get up and running in under 5 minutes:

  1. Download Speakeasy from our website and drag it to your Applications folder.
  2. Launch Speakeasy and accept the EULA and Privacy Policy.
  3. Complete the quick tour to learn the basics.
  4. Grant permissions when prompted (Microphone and Accessibility).
  5. Choose your setup - free macOS Speech, cloud APIs, or local Whisper.
  6. Start dictating! Hold Right Option or Fn and speak.

Choose Your Setup

Speakeasy offers three main ways to transcribe your voice:

Free Mode - macOS Speech Recognition

Uses Apple's built-in speech recognition. No API keys needed.

  • Completely free - no API costs
  • Works offline
  • Good accuracy for common speech
  • Can be combined with Claude Desktop for text cleanup

Cloud Mode - OpenAI/Deepgram API

Best accuracy using cloud-based AI transcription services.

  • OpenAI Whisper - $0.006/minute, excellent accuracy
  • Deepgram Nova-2 - $0.0043/minute, fast and accurate
  • AssemblyAI - $0.00025/second, good for long recordings
  • Requires internet connection

Local Mode - Whisper on Your Mac

Download Whisper models and run everything locally.

  • Completely free after model download
  • Works offline - no internet needed
  • Models range from 75MB (tiny) to 3GB (large-v3)
  • Best on Apple Silicon Macs (M1/M2/M3/M4)

API Key Configuration

If you choose cloud transcription, you'll need API keys from your chosen provider.

Service Purpose Cost Get Key
OpenAI Whisper transcription $0.006/min platform.openai.com
Deepgram Nova-2 transcription $0.0043/min console.deepgram.com
AssemblyAI Transcription $0.00025/sec assemblyai.com
Anthropic Claude text cleanup ~$0.003/request console.anthropic.com

Adding API Keys to Speakeasy

  1. Open Speakeasy Settings (Cmd + ,)
  2. Go to the General tab
  3. Paste your API keys in the appropriate fields
  4. Click Test to verify your OpenAI key works
Tip: Start with OpenAI For the best experience, we recommend starting with an OpenAI API key for Whisper transcription. It offers excellent accuracy at a reasonable cost (about $0.36 per hour of dictation).

Transcription Providers

Choose your transcription provider in Settings > Backend:

macOS Speech

Free

Built-in Apple speech recognition. Works offline, no setup needed.

OpenAI Whisper

$0.006/min

Industry-leading accuracy. Supports 50+ languages.

Deepgram Nova-2

$0.0043/min

Fast, accurate, and cost-effective. Great for real-time use.

Local Whisper

Free (offline)

Download models and run locally. Best on Apple Silicon.

Text Cleanup Options

After transcription, Speakeasy can clean up your text to remove filler words and fix grammar. Choose your cleanup provider in Settings > Backend:

None

Free

Raw transcription without cleanup. Fastest option.

Anthropic API

~$0.003/request

Claude AI for intelligent cleanup. Best for UK English.

Claude Desktop (MCP)

Free with subscription

Uses your Claude Desktop subscription via MCP.

Ollama (Local)

Free (local)

Run Llama or other models locally. Requires Ollama app.

Setting Up Claude Desktop (MCP)

If you have a Claude Desktop subscription, you can use it for free text cleanup:

  1. Install Claude Desktop if you haven't already
  2. Keep Claude Desktop running in the background
  3. In Speakeasy Settings > Backend, select "Claude Desktop (MCP)" as your cleanup provider
  4. Speakeasy will automatically detect Claude Desktop and use it for cleanup

Setting Up Ollama (Local AI)

For completely offline text cleanup:

  1. Install Ollama from ollama.ai
  2. Download a model: ollama pull llama3
  3. In Speakeasy Settings > Backend, select "Ollama (Local)"
  4. Enter your model name (e.g., "llama3")

Local Backend (Optional)

The local backend enables offline transcription using Whisper AI models running entirely on your Mac.

What is the Local Backend? The backend is a small server (14 MB) that provides transcript history storage using SQLite. It runs locally on port 10001.

When Do You Need It?

Required For:

  • Transcript history storage (searchable history of all transcriptions)

NOT Required For:

  • Local Whisper transcription (uses bundled whisper.cpp runtime)
  • macOS Speech Recognition (free, built-in)
  • OpenAI Whisper API (cloud)
  • Deepgram API (cloud)
  • Claude Desktop MCP (text cleanup)
  • Anthropic API (text cleanup)
  • Ollama (text cleanup - runs separately)

Installing the Local Backend

  1. Open Speakeasy Settings (Cmd + ,)
  2. Go to the Backend tab
  3. Find the Local Backend section
  4. Click "Install Local Backend" (14 MB download)
  5. Optionally enable "Start backend automatically"

Backend Controls

Once installed, you can:

  • Start/Stop - Manually control the backend
  • Auto-start - Launch automatically when Speakeasy opens
  • Uninstall - Remove the backend to free up space
Note: The backend includes SQLite for transcript history and runs on port 10001. It only accepts connections from localhost for security.

Offline Mode (Local Whisper)

For complete offline operation with no API costs, Speakeasy includes a bundled GPU-accelerated Whisper runtime powered by whisper.cpp.

GPU Acceleration on Apple Silicon On M1/M2/M3/M4 Macs, transcription runs on the GPU using Metal for maximum performance. Intel Macs use CPU-based transcription.

Setting Up Local Whisper

  1. Open Settings > Backend > Local Whisper
  2. Click "Install Whisper Runtime" (~4 MB)
  3. Enable "Use Local Whisper"
  4. Select a model size and click "Download Model"

Available Models

Model Size Speed Accuracy
Tiny 75 MB Fastest Good
Base 145 MB Fast Better
Small 488 MB Medium Good
Medium 1.5 GB Slower Great
Large-v3 3 GB Slowest Best

Updating Whisper Runtime

To update to the latest whisper.cpp version bundled with new app releases:

  1. Go to Settings > Backend > Local Whisper
  2. Click "Update Runtime" to reinstall from app bundle
Recommendation for Apple Silicon On M1/M2/M3/M4 Macs, the "Small" or "Medium" model offers the best balance of speed and accuracy. The "Large-v3" model provides the best quality but requires more processing time.

Required Permissions

Speakeasy needs the following macOS permissions to function:

Microphone Access

Required to record your voice for transcription.

  • Speakeasy will prompt you automatically on first use
  • Or enable manually: System Settings > Privacy & Security > Microphone

Accessibility Access

Required to paste text and detect keyboard shortcuts globally.

  • System Settings > Privacy & Security > Accessibility
  • Click the lock to make changes
  • Add Speakeasy to the list and enable it

Input Monitoring (Optional)

Improves hotkey detection reliability.

  • System Settings > Privacy & Security > Input Monitoring
  • Add Speakeasy to the list

Keyboard Shortcuts

Shortcut Action
Right Option (hold) Dictate in Quick mode
Fn (hold) Dictate in Stream mode
Left Option (hold) Dictate in Whisper mode (quiet speech)
Left Control (double-tap) Toggle Coding Mode
Cmd + K Open Coding Commands reference
Shift + Cmd + H Open Clipboard History
Cmd + , Open Settings

License Activation

Speakeasy uses a "pay what you think" pricing model. You decide what the software is worth to you.

Getting a License

  1. Visit the Lemon Squeezy store
  2. Choose your price ($0 to $99.99)
  3. Complete checkout
  4. Receive your license key via email

Activating Your License

  1. Open Speakeasy Settings (Cmd + ,)
  2. Go to the License & Legal tab
  3. Enter your license key
  4. Click Activate
7-Day Trial Speakeasy includes a 7-day trial period. You can use all features without a license key during this time. After 7 days, you'll need to activate a license (even a free $0 one) to continue using the app.

License Details

  • Duration: 1 year from activation
  • Devices: 2 device activations included
  • Updates: All updates during your license period
  • Renewal: Renew yearly at your chosen price

Troubleshooting

Hotkey not working

  • Check Accessibility permission is enabled
  • Try restarting Speakeasy
  • Try a different key (Right Option, Fn, or Left Option)
  • Some keyboards remap Fn - try Right Option instead

No audio captured

  • Check Microphone permission is enabled
  • Test your microphone in System Settings > Sound
  • Ensure the correct microphone is selected

Text not pasting

  • Check Accessibility permission is enabled
  • Some apps block automated input - use clipboard history instead
  • Try restarting the target application

API errors

  • Verify your API key is correct
  • Check you have API credits/balance
  • Ensure the key has the correct permissions
  • Test your internet connection

License validation failed

  • Check your internet connection
  • Verify the license key is entered correctly (no extra spaces)
  • Click "Check" to test license server connectivity

Need Help?

If you're still having issues: