Updated July 27, 2026
Privacy at EarScribe
EarScribe is designed so the audio file you choose is decoded and transcribed on your device. This page explains that boundary without claiming that a web page makes every privacy or security risk disappear.
Your audio and transcript
The selected audio is passed to browser audio APIs and a Whisper model running in a Web Worker. EarScribe does not upload that audio to its application server.
Transcript edits may be saved in this browser so they can be restored when you select the same file again. Audio is not included in that saved draft.
- You can clear saved drafts through your browser site-data settings.
- Exported files are created locally and downloaded by your browser.
What the browser downloads
The page loads HTML, JavaScript, CSS and fonts or images used by the interface. On first use, the selected Whisper model is downloaded through EarScribe's model proxy and may be cached by the browser.
- Model downloads range from roughly 80 MB to 800 MB.
- Removing browser cache or site data may remove cached models and saved drafts.
Aggregate analytics
EarScribe uses Plausible Analytics to understand aggregate visits and real product actions such as selecting a file, starting a transcription, completing it, using subtitle checks and exporting a format.
- Audio, transcript text and file names are not sent as analytics properties.
- Device capability is attached only to real workflow events to diagnose compatibility and performance.
Limits of local processing
Local processing reduces exposure created by uploading recordings to a transcription service, but it does not make the browser, device, extensions, network or downloaded exports automatically secure. Use device and organizational controls appropriate to the sensitivity of the recording.