Local, CPU-Friendly, High-Quality TTS (Text-to-Speech) with Kokoro

	Local, CPU-Friendly, High-Quality TTS (Text-to-Speech) with Kokoro(ariya.io)
	508 points by speckx 98 days ago \| 94 comments
	tl;dr: Kokoro is an 82M-parameter TTS model that produces high-quality multilingual speech entirely on CPU, easily deployable via the Kokoro-FastAPI Docker container which exposes an OpenAI-compatible speech API on port 8880. Benchmarks show it's fast even on old hardware—generating a short paragraph in 4.7s on a 12-year-old Intel i7-4770K and 1.5s on a Ryzen 7 8745HS. For users needing STT as well, the Speaches container bundles both Kokoro and Whisper.
	HN Discussion: ↑Users praise Kokoro's quality and CPU-friendliness, sharing successful personal use cases like ebook/article reading ~Kokoro has limitations such as poor short-word pronunciation and limited language support •Other TTS models (Chatterbox, Fish Audio, Pocket TTS) offer more features like voice cloning and emotional control ↑Developers are building their own tools and optimizations around Kokoro (browser extensions, mobile ports, games) ↑Kokoro punches above its weight compared to other TTS models per benchmarks