Multilingual speech recognition with a ternary (1.58-bit) BitNet decoder — running fully in your browser with 🤗 Transformers.js. English, Chinese, French, Italian, Korean, Portuguese, Vietnamese & more.
Speak into your microphone and watch the transcription appear live — decoded locally in ~3 s windows while you talk. Press stop to finalise.
Model: microsoft/VibeVoice-ASR-BitNet · Paper: VibeVoice-ASR-BitNet Technical Report · Reference engine: VibeASR.cpp · Server-side demo: vibevoice-asr-bitnet-demo