piper-plus

piper-plus is an MIT-licensed, fully offline neural text-to-speech engine and a maintained fork of Piper (rhasspy/piper, archived in 2025; its successor moved to GPL-3.0). It provides pretrained voices for Japanese, English, Chinese, Spanish, French and Portuguese, its own grapheme-to-phoneme (G2P) front ends for eight languages without espeak-ng, and runtimes for Python, Rust, C#, Go, C++ and WebAssembly (in the browser), plus iOS and Android G2P libraries. Features include streaming, phoneme timing output, custom pronunciation dictionaries, SSML and voice cloning with precomputed speaker embeddings. It is used by other software, including the YukkuriMovieMaker 4 video editor, and is a permissively licensed option for embedding speech synthesis in apps, games, robots and accessibility tools without sending text to the cloud.