Kokoro is an open-source TTS model (82M parameters) that runs efficiently on CPU, including Apple Silicon, without requiring a dedicated GPU. This fills a gap for users without powerful GPUs, enabling high-quality TTS on everyday computers for accessibility, reading, and automation workflows. Kokoro supports manual IPA pronunciation guides and has a CLI tool; it produces natural-sounding speech but may struggle with single words or homographs.
Background
Text-to-speech (TTS) converts written text into spoken audio. High-quality TTS often requires neural networks that benefit from GPU acceleration, making local CPU-only solutions rare. Kokoro is designed to be efficient on CPUs, leveraging the mlx-audio library on Apple Silicon.
References
Discussion
Community comments show enthusiastic adoption. Users appreciate Kokoro for accessibility products, reading articles via RSS, and even as the backbone of voice-controlled interfaces. Some note limitations with homographs and single-word utterances, but overall sentiment is positive.