MeloTTS is a high-quality, multilingual text-to-speech library from MyShell.ai, the team behind a popular AI voice platform. It synthesizes natural speech in English, Spanish, French, Chinese, Japanese and Korean — including mixed Chinese-English text — and is fast enough to run in real time on a CPU. The project has earned about 7,700 stars on GitHub, with the library last updated in late 2024.
Most open TTS systems are either English-only or too slow for interactive use without a GPU. MeloTTS was built to close that gap: its models are lightweight, inference is optimized for everyday hardware, and the voices are trained for the quality bar of a consumer voice product rather than a research demo. English support includes several regional accents (American, British, Indian, Australian and a default voice), and Chinese text with embedded English reads naturally instead of switching engines mid-sentence — a detail that matters for real content. The library ships with pretrained models, a simple Python API and speaker/style controls, and it powers voice features in MyShell’s own products, so it has been hardened by production use. For developers adding narration, accessibility or voice interfaces in multiple languages, it remains one of the most practical open options.
MeloTTS is free and open source under the MIT license, including its pretrained models.
MeloTTS is for developers who need fast, good-sounding speech synthesis in several languages — for narration, accessibility, voice bots or content production — without a GPU or a cloud bill.