Kokoro TTS

At just 82 million parameters, Kokoro is an open weight text-to-speech model that produces speech quality comparable to much larger systems while running faster and at lower cost. The accompanying Python library installs via pip and handles inference, including streaming audio output at a 24 kHz sample rate. Kokoro speaks eight languages, American and British English, Spanish, French, Hindi, Italian, Japanese, Brazilian Portuguese, and Mandarin Chinese, with multiple voices available. It depends on espeak-ng for phonemization and runs on CPU, CUDA GPUs, or Apple Silicon, making real time synthesis feasible on consumer hardware without a dedicated graphics card. Because the weights are Apache licensed, commercial deployment is permitted, and the model shows up in production voice features as well as hobby projects where cloud TTS would be too slow or too expensive. Developers building reading apps, voice assistants, and audio pipelines are typical users.