Full policy (Vietnamese): https://tolkora.com/privacy/

What Tolkora sends to its server
- The audio of the video you are watching, only on sites you have enabled and only while the video is open. The server uses it for speech recognition and then discards it; audio is not written to disk.
- Existing captions of the video, if any (translated instead of listening).
- The video id, title and channel name, to translate names consistently and to reuse finished subtitles.
- The playback position, to translate ahead of where you are and to count minutes used.
- An anonymous device identifier, generated randomly on first use. It is not tied to a name, email or account.

What the server keeps
- Translated subtitles per video (original line and translation). They do not record who watched.
- Minutes used per day, tied to the anonymous identifier, for the free allowance.
- Server logs for 14 days (anonymous identifier, video id, time), then deleted.
- A hash of the network address at first registration, to limit how many identifiers one network can create. The real address is not stored.
- A hash of the recovery code, if you create one.

What Tolkora does not do
- It reads nothing on sites you have not enabled, and does not run until you agree on the welcome page.
- It does not collect microphone input, browsing history or passwords.
- It does not sell, rent or share data with advertisers or any third party.
- It does not send audio or speech to any third-party AI service: recognition and translation run on Tolkora's own server.

Processor: connections pass through Cloudflare, which handles network addresses and traffic only to relay and protect the service.

Your choices: you can turn Tolkora off per site or remove it at any time. Removing it deletes the identifier from the browser.

Contact and deletion requests: hello@tolkora.com (include your recovery code so the right identifier can be found).