speak-tts

Voice cloning CLI for Mac (Apple Silicon) using F5-TTS.

Installation

Via Homebrew

brew tap danielsinai/speak-tts
brew install speak-tts

Via pipx

pipx install speak-tts

Via pip

pip install speak-tts

Requirements

• macOS with Apple Silicon (M1/M2/M3/M4)
• Python 3.10+

Usage

Add a Voice Profile

speak --add-profile daniel --voice ~/voice.wav --text-file ~/reference.txt

The reference audio should be a clear WAV recording (24kHz) of the voice you want to clone, and the text file should contain an exact transcript.

Generate Speech

Plays audio directly (no file saved):

speak daniel "Hello, this is a test!"

Save to a file:

speak daniel "Hello world" --output ~/Desktop/hello.wav

List Profiles

speak --list

Play Audio

speak --play output.wav

Profile Storage

Profiles stored in ~/.speak/profiles/<name>/ with:

• reference.wav - Voice sample
• reference.txt - Transcript

Tips

• Reference Audio: 5-15 seconds, 24kHz, minimal background noise
• Reference Text: Exact transcript of the audio
• First Run: Downloads F5-TTS model (~1GB)

Publishing

1. Publish to PyPI

# Install tools
pip install build twine

# Build and upload
./scripts/publish.sh

Or manually:

python -m build
twine upload dist/*

2. Set up Homebrew Tap

1. Create a new GitHub repo: homebrew-speak-tts

2. Copy the formula:

   cp Formula/speak-tts.rb ~/homebrew-speak-tts/
   

3. Get SHA256 after PyPI upload:

   shasum -a 256 dist/speak_tts-*.tar.gz
   

4. Update sha256 in the formula, commit, and push

5. Users install with:

   brew tap YOUR_USERNAME/speak-tts
   brew install speak-tts
   

Development

git clone https://github.com/danielsinai/speak-tts
cd speak-tts
python3 -m venv venv
source venv/bin/activate
pip install -e .
speak --help

License

MIT