Voice Commands

Dictate your commit messages using Gitux's built-in voice recognition.

Overview

Gitux includes a local Whisper AI model for speech-to-text. This means:

• ✅ Completely offline — No internet required after initial setup
• ✅ Private — Audio never leaves your computer
• ✅ Free — No API keys or subscriptions
• ✅ Fast — Transcription takes just seconds

First-Time Setup

The first time you use voice input:

1. Click the microphone button or press Alt+M
2. Gitux downloads the Whisper model (~40MB)
3. Wait for "Model Ready" notification
4. The model is cached for future use

One-Time Download

The AI model only downloads once. After that, voice input works instantly, even offline.

Using Voice Input

Start Recording

• Click the microphone button in the commit area
• Or press Alt+M

While Recording

• A red indicator shows recording is active
• A countdown shows remaining time (max 30 seconds)
• Speak clearly at a normal pace

Stop Recording

• Click the microphone button again
• Or press Alt+M
• Or wait for the 30-second limit

After Recording

1. "Processing..." appears while transcribing
2. Transcribed text is added to your commit message
3. Edit the text if needed before committing

Tips for Better Recognition

Speak Clearly

• Normal conversational pace
• Pronounce technical terms carefully
• Pause briefly between sentences

Environment

• Minimize background noise
• Use a decent microphone
• Avoid echo-prone rooms

Message Structure

Speak your commit message naturally:

"Add user authentication with JWT tokens"

"Fix the bug where users couldn't log out on mobile"

"Update documentation for the API endpoints"

Combining with Prefixes

Voice input works great with commit prefixes:

1. Press Alt+1 to add feat: prefix
2. Press Alt+M to start recording
3. Say: "Add shopping cart functionality"
4. Result: feat: Add shopping cart functionality

The prefix is preserved, and your voice input is appended.

Recording Limits

Limit	Value
Maximum duration	30 seconds
Minimum duration	~1 second
Audio format	Captured at 16kHz

30 seconds is plenty for most commit messages. If you need more, record in multiple segments.

Troubleshooting

"Microphone Error"

• Check that your browser/app has microphone permissions
• Ensure your microphone is connected and working
• Try a different microphone

"No Speech Detected"

• Speak louder or closer to the microphone
• Check that your microphone is not muted
• Reduce background noise

Model Won't Load

• Check your internet connection (first time only)
• Ensure you have ~100MB free disk space
• Restart Gitux and try again

Poor Transcription Quality

• Speak more slowly and clearly
• Reduce background noise
• Try a better microphone

Privacy & Security

Local Processing

The Whisper model runs entirely on your CPU. Your audio is:

• Processed in memory
• Never saved to disk
• Never transmitted anywhere

No Cloud Services

Unlike other voice tools, Gitux doesn't use:

• ❌ Google Speech API
• ❌ Amazon Transcribe
• ❌ OpenAI Whisper API
• ❌ Any cloud service

Everything happens on your machine.

Technical Details

Model

Gitux uses whisper-tiny optimized for:

• Fast loading (~2 seconds)
• Quick transcription (~3 seconds for 10s audio)
• Small size (~40MB)

Supported Languages

The model works best with English but supports multiple languages. Accuracy varies by language.

System Requirements

• Any modern CPU (no GPU required)
• ~100MB disk space for model
• Working microphone

Keyboard Reference

Shortcut	Action
Alt+M	Toggle voice recording

Workflow Example

1. Stage your files (Ctrl+S)
2. Add prefix (Alt+2 for fix:)
3. Start recording (Alt+M)
4. Say: "Resolve the timeout issue on slow connections"
5. Stop recording (Alt+M)
6. Review: "fix: Resolve the timeout issue on slow connections"
7. Commit (Ctrl+Enter)

Next Steps

• Stashing Changes — Save work temporarily
• File History — Track file changes over time