Documentation

System Requirements

• Operating System: Windows 11 or later (64-bit)
• RAM: 4 GB minimum (8 GB recommended for local transcription)
• Storage: 200 MB for app + optional model downloads (31-190 MB for Q5_1 models)

Installation

1. Download the installer from the download page
2. Run Finch_x.x.x_x64-setup.exe
3. Follow the installation wizard
4. Finch will launch automatically after installation

First Launch

On first launch, you'll see a 3-step onboarding flow:

1. Welcome: Introduction to Finch
2. Transcription Setup: Choose between Local (private, offline) or Cloud (Groq - fast)
3. Ready: Your hotkey is set to Ctrl + ` by default

The Minibar

After setup, you'll see a small floating button at the bottom of your screen. This is the Minibar:

• Left-click: Start/stop recording (same as hotkey)
• Right-click: Open settings menu
• Recording indicator: Button pulses red while recording

Finch offers three recording modes to match your workflow:

Push-to-Talk (Default)

Hold your hotkey while speaking. Release to transcribe.

• Best for: Quick dictation, avoiding accidental recordings
• How to use: Hold Ctrl + `, speak, release
• Tip: Great for inserting short phrases or commands

Toggle Mode

Press hotkey once to start recording, press again to stop.

• Best for: Longer dictation sessions, hands-free operation
• How to use: Press Ctrl + ` to start, press again to stop
• Tip: Ideal for dictating paragraphs or documents

Voice-Activated (VAD)

Automatically detects when you start and stop speaking.

• Best for: Completely hands-free dictation
• How to use: Press hotkey to activate, then just speak naturally
• Auto-stop: Recording ends after detecting silence
• Tip: Works best in quiet environments

Changing Recording Mode

1. Right-click the Finch button
2. Select Settings
3. Go to the Recording section
4. Choose your preferred mode

Finch supports multiple transcription engines. Choose based on your needs:

Local (Whisper)

Process audio entirely on your device using OpenAI's Whisper model.

• Privacy: Audio never leaves your computer
• Offline: Works without internet
• Cost: Free forever

Finch uses quantized Q5_1 models optimized for speed and small download size.

Cloud: Groq Zero Data Retention

Fast cloud processing using Groq's Whisper API.

• Speed: Fastest transcription (sub-second)
• Privacy: Zero data retention by default - audio is not stored
• Cost: Free tier available
• Setup: Requires API key from console.groq.com

Cloud: Deepgram Zero Data Retention SOC 2 Type II

Real-time streaming transcription - see text as you speak.

• Real-time: Words appear as you say them
• Privacy: Zero storage by default - audio is processed and immediately discarded
• Cost: Free tier with 200 hours
• Setup: Requires API key from console.deepgram.com

Finch uses AI to enhance your dictated text beyond simple transcription.

LLM Text Processing

When enabled, your transcribed text is processed by an AI language model to:

• Fix grammar: Correct grammatical errors
• Add punctuation: Proper capitalization and punctuation
• Clean filler words: Remove "um", "uh", "like", etc.
• Format naturally: Make text flow more naturally

Requires a Groq API key (same key used for cloud transcription).

Context Detection

Finch can detect your current context to format text appropriately:

• IDE Detection: Formats as code comments when in VS Code, IntelliJ, etc.
• App Awareness: Adapts formatting based on active application

Smart Formatting

Toggle smart formatting to automatically apply context-aware formatting:

1. Go to Settings → AI Features
2. Toggle Smart Formatting

Disabling AI Processing

You can disable AI processing for raw transcription only:

• Turn off LLM Processing in Settings → AI Features
• Or enable Privacy Mode to disable all cloud features

Control your computer with voice commands. Enable Command Mode in Settings.

Built-in Keyboard Commands

Say these phrases to trigger keyboard shortcuts:

Custom AI Commands

Create your own voice-triggered AI actions:

1. Go to Settings → Voice Commands
2. Click Add Command
3. Set a trigger phrase (e.g., "summarize this")
4. Choose an action type:
   • Transform: Modify selected text
   • Generate: Create new text
   • Execute: Run a custom prompt
5. Write your AI prompt

Enabling Command Mode

1. Go to Settings → Voice Commands
2. Toggle Command Mode on
3. Commands will be detected automatically during dictation

Personal Dictionary

Teach Finch words it might mishear - names, technical terms, jargon:

Adding Dictionary Entries

1. Go to Settings → Dictionary
2. Click Add Entry
3. Enter the word as spoken (e.g., "john smith")
4. Enter the correct spelling (e.g., "John Smith")
5. Optionally assign a category for organization

Example Dictionary Entries

Snippets

Snippets are voice-triggered text expansions - say a short phrase, insert longer text:

Creating Snippets

1. Go to Settings → Snippets
2. Click Add Snippet
3. Set a trigger phrase (e.g., "my email")
4. Enter the expanded text (e.g., "[email protected]")

Example Snippets

Privacy Mode guarantees zero network connections for maximum security.

What Privacy Mode Does

When enabled, Finch operates in complete isolation:

• Local-only transcription: Forces use of downloaded Whisper model
• No LLM processing: Disables AI text cleanup
• No AI commands: Only keyboard commands work
• No update checks: Automatic updates disabled
• No history: Transcripts aren't saved

Visual Indicator

When Privacy Mode is active, you'll see a shield icon on the minibar and in settings.

Enabling Privacy Mode

1. Go to Settings → General
2. Toggle Privacy Mode on
3. Requires a local model to be downloaded first

Use Cases

• Sensitive work: Legal, medical, financial dictation
• Air-gapped systems: Computers without internet access
• Corporate compliance: When data can't leave premises
• Personal preference: Maximum privacy for any reason

Finch implements multiple layers of security to protect your data at rest and in transit.

Database Encryption

All local data is stored in an encrypted SQLCipher database:

• SQLCipher: 256-bit AES encryption for dictionary, snippets, history, and custom commands
• DPAPI (Windows): Database encryption key protected by Windows Data Protection API, bound to your user account
• Auto-Migration: Existing unencrypted databases are automatically migrated on app update
• Data Location: %APPDATA%\Finch\finch.db

Credential Storage

API keys are stored securely using your operating system's credential manager:

• Windows: Credential Manager with DPAPI-protected fallback
• macOS: Keychain (when available)
• Keys are never stored in plain text files
• Keys never leave your device (except to their respective API providers)
• Never included in backup exports

Input Protection

Finch protects against malicious input patterns:

• Character Limits: 10,000 character maximum for text input to LLM processing
• Audio Limits: 100 MB maximum audio file size prevents memory exhaustion
• Prompt Injection Prevention: Code blocks, instruction patterns, and special tokens are neutralized before LLM processing

Network Security

• Signed Updates: Ed25519 cryptographic signatures verify every update
• SHA256 Checksums: Model downloads verified against known hashes from HuggingFace
• Content Security Policy: Strict CSP prevents XSS and unauthorized scripts
• HTTPS Only: All network requests use encrypted TLS connections
• Minimal Permissions: Tauri 2.0 capabilities limit app permissions to only what's needed

Third-Party Security

Cloud providers used by Finch maintain strong security practices:

• Groq: Zero Data Retention - audio/text not stored after processing
• Deepgram: Zero Data Retention + SOC 2 Type II certified
• HuggingFace: Model downloads only, no user data sent

Accessing Settings

Right-click the Finch button → Select Settings

Hotkey Configuration

1. Go to Settings → General
2. Click Change Hotkey
3. Press your desired key combination
4. Default: Ctrl + ` (backtick)

Microphone Selection

1. Go to Settings → Recording
2. Select your preferred microphone from the dropdown
3. Finch remembers your selection

Autostart

Launch Finch automatically when Windows starts:

1. Go to Settings → General
2. Toggle Start with Windows

Sound Effects

Audio feedback for recording start/stop:

1. Go to Settings → General
2. Toggle Sound Effects

Export & Import

Backup your settings, dictionary, and snippets:

• Export: Settings → Data → Export All Data
• Import: Settings → Data → Import Data
• Creates a JSON file with all your customizations

Auto-Updates

Finch checks for updates automatically:

• Update notifications appear when a new version is available
• Choose to install now or later
• Disabled when Privacy Mode is active

Microphone Not Detected

• Check Windows sound settings - ensure mic is enabled
• Try a different USB port if using USB mic
• Restart Finch after connecting new microphone
• Check microphone permissions in Windows Settings → Privacy

Hotkey Not Working

• Check if another app is using the same hotkey
• Try a different key combination
• Run Finch as administrator (right-click → Run as administrator)
• Restart Finch

Transcription Errors

• Garbled text: Speak closer to microphone, reduce background noise
• Wrong words: Add corrections to Dictionary
• Slow transcription: Try a smaller model or cloud engine
• API errors: Check your API key is valid and has credits

App Won't Start

• Check Task Manager for existing Finch process - end it
• Run as administrator
• Reinstall from the download page
• Check Windows Event Viewer for error details

Model Download Issues

• Check internet connection
• Ensure sufficient disk space (up to 190 MB for Quality mode)
• Try Speed mode first (only 31 MB)
• Check if firewall is blocking downloads from HuggingFace

Text Not Pasting

• Ensure the target application is in focus
• Some apps block automated paste - try a different app
• Check if clipboard is working (try manual Ctrl+V)