# Kitten TTS

> Apache-2.0 Python TTS library that runs on a CPU

**Category:** Python  
**Pricing:** open-source  
**URL:** https://kittenml.com  
**GitHub:** https://github.com/KittenML/KittenTTS  
**Added:** 2026-10-16

---

## The Hook

Kitten TTS is a pip-installable text-to-speech library: pip install kittenml, load a model, and generate 24 kHz speech locally without a GPU or a cloud voice API.

## What's Good

Apache-2.0 code (KittenML/KittenTTS; kittenml 0.9.4 on PyPI). The default KittenTTS 2 model (1.7B parameters, 1-bit) has 47 built-in voices across 20 languages, clones a voice from 5–30 seconds of audio and takes [emotion] tags and inline events. The lightweight 0.8 ONNX models (15M–80M parameters, 25–80 MB) run on a plain CPU, and both families load through the same KittenTTS(...) call. Long text is chunked by sentence, and numbers, dates and units are expanded for you. ★~16k.

## The Catch

The weights are licensed separately from the code. KittenTTS 2 uses the Stellon Labs Community License: free for research and non-commercial use, and for commercial use only until you or your affiliates pass USD 1M in annual revenue or total funding. The small 0.8 ONNX models are Apache-2.0. KittenTTS 2 wants about 6 GB of RAM on CPU or about 8 GB on a GPU, emotion control is beta, and only clone voices you have consent to use.

## Verdict

A strong local TTS default for agents and apps — use the Apache-2.0 ONNX models if the KittenTTS 2 license is a problem for you.