Live demo
Neural TTS, in your browser
This is the same Supertonic-3 model that powers react-native-tts-kit, running here via ONNX Runtime Web — WebGPU when your browser has it, WebAssembly otherwise. 31 languages, 10 voices, and after the one-time model download it works with the network unplugged.
The multilingual Supertonic-3 model is a one-time 213 MB download from HuggingFace, cached by your browser. Synthesis then runs entirely on this page — your text never leaves it.
Model: Supertonic-3 by Supertone Inc. (BigScience OpenRAIL-M — AI-generated audio must be disclosed; no impersonation without consent). Read the launch post: on-device neural TTS in three lines.