eSpeak NG

eSpeak NG is a compact open-source speech synthesizer that supports more than 100 languages and accents. Instead of neural networks it uses formant synthesis, which keeps the whole system down to a few megabytes and keeps speech intelligible even at very high reading speeds, at the cost of a distinctly robotic timbre. Written in C, it runs on Linux, Windows, Android, macOS, and BSD, and works as a command-line program, a shared library, or a Windows SAPI5 voice; it can also act as a front end to MBROLA diphone voices for more natural output. SSML and HTML markup are supported, and the C API stays compatible with the original eSpeak. Because it needs no GPU and almost no resources, it is a staple of screen readers such as NVDA, embedded devices, and accessibility projects, and its text-to-phoneme rules are often borrowed as a front end by neural TTS systems. Licensed under GPL 3.0.