Space SpeakP Play Audio→ SkipE ExampleH PinyinS Shuffle
⚙️ Speech & Engine Settings
Sherpa-ONNX WebAssemblyOn-Device AI
100% offline neural recognition running locally on your device. Zero cloud dependency, zero audio transmitted over the network. Silero voice-activity detection is always on: it segments your speech automatically for as long as the microphone is live.
80%
Percentage phonetic similarity required to advance. Lowering to 70% is more forgiving of slight accent/tone variations.