CatacoLabs Kokoro TTS CLI

A small, local-first CLI for the open-weight Kokoro TTS model, optimized for macOS/Linux.

Goals:

• dead-simple CLI (stdin → wav)
• keeps weights out of git (download on demand)
• works great on Apple Silicon CPU; optional server/warm mode later

Quickstart (macOS)

0) Use Python 3.12/3.13

Kokoro ONNX tooling currently targets Python < 3.14. On this Mac mini, the simplest path is:

brew install python@3.12

1) Create a venv + install

/opt/homebrew/bin/python3.12 -m venv .venv
source .venv/bin/activate
pip install -e .

2) Download model files (one-time)

kokoro-cli download

This downloads into ~/.cache/kokoro-cli/ by default.

3) Synthesize

echo "Hello Luis, local TTS is live." | kokoro-cli say --voice af_sky --out out.wav

Play it on macOS:

afplay out.wav

Commands

• kokoro-cli download — downloads kokoro-v1.0.onnx + voices-v1.0.bin
• kokoro-cli voices — list available voices
• kokoro-cli say — one-shot synthesis (loads model per run)
• kokoro-cli serve — daemon mode (keeps model loaded)
• kokoro-cli client — call the daemon (fast)

Daemon mode (recommended)

Terminal 1:

kokoro-cli serve --host 127.0.0.1 --port 7777

Terminal 2:

echo "instant TTS" | kokoro-cli client --voice af_sky --out out.wav --play

Model sources

This tool uses the community ONNX packaging for Kokoro:

• kokoro-onnx (Python lib): https://github.com/thewh1teagle/kokoro-onnx

We download model artifacts at runtime and do not commit them.

License

Apache-2.0 (repo code). Model and voice files retain their upstream licenses.