Privacy guide

How to transcribe audio without uploading it

Local transcription keeps the selected media on your device, but the website and model still use the network. Here is the complete boundary and a practical review workflow.

Updated August 30, 2026
Short answer

Use a tool that decodes the file and runs its speech model on the current device. Whisper Web keeps the selected media in browser memory; only the site and model files are downloaded.

Choose a recording
01

A no-upload transcription workflow

  1. 1

    Choose a local-processing tool

    Confirm that inference runs in the browser or on the device, not on a remote transcription API.

  2. 2

    Select only the intended file

    Whisper Web can read a chosen file, a microphone recording or a public direct media URL that allows browser access.

  3. 3

    Wait for local model processing

    The first run downloads a Whisper model. Later runs can reuse the browser cache until it is cleared.

  4. 4

    Review and export

    Correct the transcript locally and export only the format needed for the next step.

02

What stays local and what still uses the network

ItemWhere it goes in Whisper Web local mode
Selected file or microphone recordingDecoded and processed on the current device
Whisper modelDownloaded to the browser, then used locally
Application code and fontsDownloaded when the site loads
Completed transcriptStored in this browser's IndexedDB until exported or deleted
Direct media URLFetched by the browser from the source server; CORS must allow access
03

Check the whole workflow, not just the upload button

A no-upload claim should explain model downloads, browser storage and direct URL behavior. It should also say whether analytics, crash logs or cloud features are part of the same page.

Local processing reduces exposure to a transcription provider, but device backups, shared accounts and exported files can still copy sensitive content elsewhere.

04

Questions about this guide

Does no upload mean the website works fully offline?

No. The application and selected Whisper model must download first. The current product does not promise a fully offline website session.

Where is the completed transcript stored?

Whisper Web stores it in IndexedDB in the current browser until you export or delete it.

Is a pasted media URL private?

The browser must request that URL from its source server. Use a URL only when you understand the source server's access and logging behavior.

Transcribe on this device

Choose a supported audio or video file.

Choose a file