Privacy Policy — Local Live Captions

Last updated: 2026-09-13

Local Live Captions does not collect, transmit, or sell any personal data. There is no analytics, no telemetry, and no account.

What the add-on processes

Audio from the tab or your microphone. It is read only while captions are on, converted to 16 kHz mono in memory, transcribed, and discarded. It is never written to disk and never sent anywhere in the default (local) mode.
Captions. Shown on screen only. They are not stored, logged, or uploaded, and are discarded when the panel closes.
Your settings (model, appearance, panel position and size, timing). Stored locally with the browser’s extension storage. They never leave your computer.
Network connections

When Where What is sent
First use of a speech model, once per model huggingface.co and its CDN A normal file download of the model weights and its tokenizer — data files, not program code. No audio, no page data, no identifiers beyond what any HTTPS request includes.
Only if you explicitly configure a remote endpoint in Settings The URL you enter 16 kHz WAV audio segments and, if you set one, your API key. This is off by default; in local mode it never happens.
No other connections are made. The add-on does not talk to any server operated by the developer, because there isn’t one.

Permissions

Access to the sites you visit — required to read the audio of pages you play media on and to draw the caption panel there. Firefox asks you to grant this explicitly, and you can grant it per site.
Microphone — only when you switch the audio source to Microphone; Firefox asks each site for permission in the usual way.
Storage — your settings.
Tabs — to know which tab is playing and to keep captions per tab.
Contact

Questions or reports: open an issue at https://github.com/EricZhou866/LiveCaption/issues