Privacy

Speech recognition and translation run on your PC. Nothing is sent online unless you turn on Cloud AI providers and pick an online provider as the LLM server.

• No online fallback. If the local model can't be reached, a line is shown untranslated; it isn't sent anywhere else.
• With an online provider picked, the recognised text of each line (with the previous lines as context) — or, with a Qwen LiveTranslate model, the tab's audio — is sent to that provider under its terms. The popup's footer says Cloud AI providers on while the setting is on. See Cloud providers & API keys, including how API keys are kept and cleared.
• The app server only accepts the extension. It only listens on your PC (127.0.0.1) and refuses connections from web pages, so a site open in your browser can't use it or read what it reports.
• Transcripts are off by default. When you turn them on, they're saved as files in the app's transcripts/ folder on your PC — nowhere else.
• Profiles and exports never contain API keys.
• Model downloads for imported profiles: Find and download looks a model up on Hugging Face by its name, publisher and quantization — only those are sent, and only when you press the button.