Convert your voice to text instantly. Press a key to start recording, press again to stop—your words appear as typed text in any application.

• Simple: One keyboard shortcut to start/stop recording
• Smart: Automatically removes silence and pauses from your recording
• Fast: Text appears in seconds via your active window
• Clean: Removes "um", "uh" and repetitions automatically
• Multilingual: Supports Hindi with Roman script output (Hinglish)
• Lists: Detects and formats bullet points automatically
• Safe: Your clipboard is preserved—won't lose what you copied
• Efficient: Only processes speech, ignoring silence to save costs

# Run directly without installing
nix run github:truroshan/nixspr -- start
nix run github:truroshan/nixspr -- process

# Install to your profile
nix profile install github:truroshan/nixspr

# Or add to your NixOS configuration
{
  inputs.nixspr.url = "github:truroshan/nixspr";

  # Then in your configuration:
  environment.systemPackages = [
    inputs.nixspr.packages.${system}.default
  ];
}

# Setup environment (if using direnv)
cp .envrc.example .envrc
# Edit .envrc with your API key
direnv allow

# Or manually set environment variables
export NIXSPR_GEMINI_API_KEY="your-api-key-here"

# Enter development shell
nix develop

# Run the script
python3 nixspr.py start
python3 nixspr.py process

# Or build and run
nix build
./result/bin/nixspr start

1. Obtain a Gemini API key from Google AI Studio

2. Add your API key:

Option A: For keyboard shortcut use (recommended) Add to your window manager config (see Keyboard Shortcuts section below)

Option B: For command-line use Add to your shell profile (~/.bashrc or ~/.zshrc):

export NIXSPR_GEMINI_API_KEY="your-api-key-here"

Option C: Just testing Pass it directly when needed:

nixspr process --gemini-api-key "your-api-key-here"

# 1. Start recording
nixspr start

# 2. Speak your text

# 3. Stop and convert to text
nixspr process

# Your text will appear where your cursor is!

# Allow longer recordings (up to 2 minutes)
nixspr process --max-audio-duration 120

# Cancel anytime
nixspr cancel

# Troubleshooting mode
nixspr process --debug

# One-time API key (without setting environment variable)
nixspr process --gemini-api-key "your-key"

# Use different AI model
nixspr process --gemini-model "gemini-2.5-pro"

• Maximum: 60 seconds of actual speaking (pauses don't count)
• No minimum: Automatically detects if you actually spoke
• If you exceed the limit, you'll get a notification
• Silence and pauses are removed automatically—only your speech counts

Set up a convenient keyboard shortcut for hands-free operation:

• Alt+Space: Start/stop recording
• Alt+Shift+Space: Cancel

Example for niri window manager (~/.config/niri/config.kdl):

// Add your API key here
environment {
    NIXSPR_GEMINI_API_KEY "your-api-key-here"
}

binds {
    // Press once to start, again to convert to text
    Alt+Space { spawn "sh" "-c" "[ -f /tmp/nixspr.pid ] && nixspr process || nixspr start"; }

    // Cancel anytime
    Alt+Shift+Space { spawn "nixspr" "cancel"; }
}

1. Press shortcut → Recording starts
2. Speak naturally → Audio is captured
3. Press shortcut again → Processing begins:
   • Silence and pauses are removed
   • Your speech is converted to text by AI
   • Filler words ("um", "uh") are cleaned up
   • Text is formatted with punctuation
   • Result appears in your active window

Clean Speech:

• Removes "um", "uh", "like" automatically
• Eliminates repeated words
• Adds proper punctuation

Smart Formatting:

• Detects and creates bullet point lists
• Proper capitalization
• Handles mixed English and Hindi

Hindi Support:

• Converts Hindi speech to Roman script
• Example: Speaking "मुझे किताब चाहिए" becomes "mujhe kitaab chahiye"
• Works with mixed Hindi-English (Hinglish)

• Linux with Wayland display server (most modern Linux desktops)
• Audio input (microphone or headset)
• Internet connection (for AI processing)
• API key from Google AI Studio (free to get)

All software dependencies are automatically managed by Nix—nothing to install manually!

Set up your API key using one of these methods:

• Add to your shell: export NIXSPR_GEMINI_API_KEY="your-key"
• Add to window manager config (see Keyboard Shortcuts section)
• Pass directly: nixspr process --gemini-api-key "your-key"

Check your microphone:

pactl list sources short
pactl set-default-source <your-microphone-name>

• Make sure your microphone is working
• Speak louder or closer to the microphone
• Check if other apps can hear your mic

• Default limit: 60 seconds of actual speech (pauses don't count)
• For longer recordings: nixspr process --max-audio-duration 120

• Make sure the target window is active (clicked/focused)
• Try closing and restarting the application
• Test clipboard: echo "test" | wl-copy then paste manually

Run nixspr cancel to stop everything and try again

Run with --debug to see detailed information:

nixspr process --debug

MIT