Whisper Transcriber

Web utility — Private speech-to-text in your browser

Reactions

Speech-to-textWhisperOffline AISubtitles

The first transcription downloads the model from the Hugging Face Hub and caches it in the browser — about 41 MB on phones, 77 MB on the desktop GPU path, and the utility shows the figure for your device before you start. Later runs work from that cache, but clearing site data or browser storage pressure can evict it and trigger a fresh download.

Key actions

Audio and video file upload, or record-then-transcribe from the microphone
Windowed conveyor pipeline for long recordings — the model stays loaded
Automatic language detection plus explicit English and Russian modes
WebGPU acceleration with an automatic WASM fallback
Runs on phones: smaller model, tighter windows, screen kept awake
Editable transcript with timestamped segments
Copy, TXT, SRT, and WebVTT export

Discussion

Comments

Loading comments...

Related tools

Comments