Start the demo and say anything in English
VoxRT streaming ASR runtime transcribes your speech in real time. Your microphone streams to a small demo server running the same engine you'd embed on a device.
How it works: your browser captures 16 kHz mono audio and streams it over a secure WebSocket to a demo server, where the VoxRT streaming ASR engine — the same on-device runtime you'd embed in your product — transcribes it and streams text back as you speak (80 ms chunk granularity, English model, ~61 MB fp16). Audio is processed in memory for transcription only and is never stored. Note this is the one VoxRT demo that leaves the page: on your own hardware the runtime runs fully offline, with no server involved. The demo server: built from the open-source voxrt-asr-linux SDK and run, for demo purposes, on a single small cloud VM — 4 GB RAM, Ubuntu 24.04 LTS x64, in New York — with one engine instance per session and a cap of 10 concurrent sessions. If it's at capacity (or you're far from NYC and see extra latency), that's the demo box, not the engine: on-device there is no network in the loop. Requirements: microphone permission and any modern browser.