LAB 003 · Speech recognition

Transcribe audio without uploading it.

Choose a short English recording, download Whisper Tiny once, and generate the transcript inside this browser.

Selected runtimeChecking…
First download
Calculating
Maximum clip
60 seconds
API key
None

onnx-community/whisper-tiny.en

Private transcription

CPU · local
First run: approximately 45 MBThe q8 CPU route downloads less but can run much more slowly.

Choose a short English audio file. It will remain on this device.

01

Decode

The browser opens the chosen audio file and resamples it to 16 kHz locally.

02

Transcribe

Whisper Tiny processes the waveform using WebGPU or the CPU fallback.

03

Review

The transcript remains editable on-screen and should be checked for mistakes.