Voice Input

Dictate commit messages using Gitux's built-in, offline AI.

Overview

Gitux includes OpenAI's Whisper model running locally on your machine:

Feature	Details
Privacy	Audio never leaves your computer
Cost	Free, no API keys needed
Internet	Not required (after first setup)
Speed	~3 seconds for 10s of audio

How It Works

1. Click the microphone button (or Alt+M)
2. Speak your commit message
3. Click again to stop
4. Text appears in the commit input

First-Time Setup

On first use, Gitux downloads the Whisper model:

1. Click the microphone button
2. "Loading Voice Model" notification appears
3. Wait ~30 seconds for download (~40MB)
4. "Model Ready" confirms success

The model is cached locally — subsequent uses are instant.

Using Voice Input

Start Recording

• Click 🎤 button in commit area
• Or press Alt+M

While Recording

• Red indicator pulses
• Countdown shows time remaining
• Maximum 30 seconds

Stop Recording

• Click 🎤 again
• Or press Alt+M
• Or wait for timeout

After Recording

• "Processing..." appears briefly
• Transcribed text added to message
• Edit if needed before committing

Combining with Prefixes

Voice input works seamlessly with commit prefixes:

1. Press Alt+1 (adds "feat: ")
2. Press Alt+M (start recording)
3. Say: "Add user profile page"
4. Press Alt+M (stop)
5. Result: "feat: Add user profile page"

Tips for Best Results

Speak Clearly

• Normal conversational pace
• Enunciate technical terms
• Brief pauses between sentences

Good Environment

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

Natural Language

Speak your message naturally:

• ✅ "Fix the bug where users can't log out"
• ✅ "Add validation to the signup form"
• ❌ "Fix colon bug colon users can't log out"

Privacy Details

What Happens to Your Audio

1. Captured by your microphone
2. Processed in memory by Whisper
3. Converted to text
4. Audio immediately discarded

What's NOT Sent Anywhere

• ❌ Audio recordings
• ❌ Transcribed text
• ❌ Any telemetry

No Cloud Services

Gitux doesn't use:

• Google Speech API
• Amazon Transcribe
• OpenAI API
• Any external service

Technical Specs

Spec	Value
Model	whisper-tiny
Size	~40MB
Load time	~2 seconds
Transcription	~3s for 10s audio
Max recording	30 seconds
Sample rate	16kHz

Troubleshooting

"Microphone Error"

• Check browser/app permissions
• Verify microphone is connected
• Try a different microphone

"No Speech Detected"

• Speak louder/closer to mic
• Check mic isn't muted
• Reduce background noise

Model Won't Load

• Check internet (first time only)
• Ensure ~100MB free space
• Restart Gitux

Poor Accuracy

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

Keyboard Shortcut

Shortcut	Action
Alt+M	Toggle voice recording

Related

• Voice Commands Guide — Detailed usage
• Staging Panel — Commit workflow
• Keyboard Shortcuts — All shortcuts