Skip to main content

Installation

Requirements​

  • Python 3.8 or later.
  • Linux, macOS, or Windows.
  • 25–80 MB for one model, plus Python dependencies and cache metadata.
  • Network access to Hugging Face on first load unless the cache is already populated.

Use a virtual environment to isolate ONNX Runtime and phonemizer dependencies.

python -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip

On Windows PowerShell, activate with .venv\\Scripts\\Activate.ps1.

Install release 0.8.1​

pip install https://github.com/KittenML/KittenTTS/releases/download/0.8.1/kittentts-0.8.1-py3-none-any.whl

The wheel installs onnxruntime, numpy, soundfile, huggingface_hub, phonemizer, and espeakng_loader.

Verify the install​

python -c "import kittentts; print(kittentts.__file__)"

Then load the smallest model:

from kittentts import KittenTTS

tts = KittenTTS("KittenML/kitten-tts-nano-0.8-int8")
print(tts.available_voices)

The first construction downloads config.json, the ONNX model, and voices.npz into the Hugging Face cache. Later constructions reuse those files.

Install from source​

Use source installation only when testing unreleased changes:

git clone https://github.com/KittenML/KittenTTS.git
cd KittenTTS
pip install -e .

Pin a commit in reproducible environments because the repository is a developer preview.

Last updated on