Kokoro

Kokoro is an open-source text-to-speech model with only 82 million parameters, making it one of the most efficient high-quality TTS models available. Despite its small size, Kokoro produces natural, expressive speech across multiple voices and styles. It runs locally on consumer hardware without a GPU, making it ideal for privacy-conscious applications, local assistants, and edge deployments. Available on Hugging Face and through the hexgrad GitHub repo.