Kokoro 82M, a compact text-to-speech (TTS) model with just 82 million parameters, is proving that size isn’t everything in speech synthesis. Developed to run entirely on local hardware, this ...