Four ways to try our CPU-first speech-to-text before you write a line of code: in your browser, in the cloud, on your desktop and on Android.
Runs entirely inside your browser. The open community model is compiled to WebAssembly and executes on your own CPU. No server, no upload, no install: speak into your mic and words appear in real time, fully offline once loaded.
Try the hosted commercial models right here in your browser. Both streaming (live partials over WebSocket) and pre-recorded (upload a file, get a final transcript with timestamps), with lower latency and higher accuracy than the community model.
A private, local-first desktop app, and a CPU-powered alternative to Wispr Flow and MacWhisper. Beyond dictation into any app, it generates movie & video subtitles, transcribes meeting recordings, and produces summaries, all processed on your own machine with nothing uploaded.
Install Kroko ASR Model Explorer from Google Play and run speech-to-text natively on your phone. Record or upload audio, compare model packs by size and accuracy, all processed on-device.
Sign up for the free trial and try Kroko on the hosted API or the on-premise server. Same models, your choice of deployment.