WASM Streaming Speech Recognition

Transcribe audio from your microphone in real time in the browser using Rust/WASM. This demo runs entirely offline on your CPU after downloading a ~950 MB model. It understands English and French, and uses the Kyutai STT model together with a WASM runtime built in Candle.

Status: Click "Start Speech" to begin

CPUs: -, Real-time factor: 0.00x, Duration: 0.0s

Transcription:

Click "Start Speech" to begin transcription

💡 This demo shows offline transcription in your browser. For more accurate cloud transcription and real-time LLM grammar correction, check out Voice Writer.