On-Device Voice Test
backend: —
models: not loaded
STT
Web Speech (Google)
Moonshine
Whisper-tiny
Live captions
TTS
Buffered
Streaming
Streaming (fine)
Streaming (half-sentence)
Backend
Auto
Force WASM
Force WebGPU
TTS precision
fp32
fp16
Reload TTS
Speech detection sensitivity
0.70
Default: 0.70
Minimum speech length to interrupt
416 ms
Default: 416 ms
Silence to close a segment
700 ms
Default: 700 ms
Pause after sentence
250 ms
Default: 250 ms
Pause after comma
100 ms
Default: 100 ms
Pause after word split
50 ms
Default: 50 ms
Pause to respond (send delay)
1200 ms
Default: 1200 ms
00:00
Tap the mic to load models
●
Conversation
Session Log
Copy