Privacy Policy

• Audio from the active tab. When you start a session, the extension captures the audio of your active tab using Chrome's tabCapture API and streams it to our server in real time so the AI can hear what is playing. The raw audio is never written to disk on our servers; it is converted to text by speech-to-text in real time and then discarded.
• The active tab's URL and page title. Used to give the AI minimal context (so it can refer to what you are watching or listening to) and to detect when you navigate somewhere new.
• Playback state events. When the page exposes a primary <video> or <audio> element, the extension observes play, pause, seek, and current playhead position so commentary stays in sync.
• A randomly-generated install identifier. On first use, the extension generates a random UUID and stores it locally in chrome.storage.local. It is sent with each new session so the server can group sessions from the same install (for example, to associate them with your account if you sign up later). It contains no personal information and is not linked to your Google account, your browsing history, or any device fingerprint.

The extension does not access your camera, your location, your browsing history, or any tabs other than the one you started a session on.

For each session, our server may persist:

• An anonymous session ID (UUID), the LiveKit room identifier used for that session, the active tab's URL, the page title, and timestamps.
• The randomly-generated install identifier described above, so multiple sessions from the same install can be grouped together (and, if you later create an account, associated with it).
• Transcripts of the tab audio and of the AI's commentary, used to maintain conversational context during the session.
• Lightweight non-personal metadata derived from the session, such as which AI persona was speaking and an internal label describing the comedic angle of a given line.
• Rolling AI-generated summaries of the conversation so the AI can stay coherent across long sessions.

All of the above is pseudonymous. We do not collect or store names, email addresses, IP addresses for profiling, account credentials, payment information, browsing history outside the active session, or any other personally identifying information. The install identifier lets us group sessions from the same browser install together, but it is not linked to a real-world identity unless you choose to create an account in the future.

To generate commentary, the server forwards the minimum data required to the following providers:

• LiveKit, real-time audio and video transport between the extension and the AI agent.
• Groq, speech-to-text on the tab audio and language-model generation of commentary.
• ElevenLabs, text-to-speech for the hosts' voices.
• LemonSlice, animated avatar rendering driven by the synthesized voices.

Each provider receives only what it needs to perform its function. We do not sell user data, and we do not share data with third parties for advertising, analytics, or any purpose unrelated to generating commentary.

The extension requests the following Chrome permissions, all used solely to provide the commentary feature:

• tabCapture, capture the active tab's audio.
• activeTab, read the URL and title of the tab you started a session on.
• scripting, inject the play/pause/seek monitor into the active page.
• sidePanel, render the avatars and controls in Chrome's side panel.
• storage, store a randomly-generated install identifier locally so sessions from the same install can be grouped together (see “What we collect from your browser” above).
• tabs, detect tab and URL changes so the session stays in sync.
• Host access (<all_urls>), required because users watch and listen to media on arbitrary websites (streaming services, podcasts, news sites, social media, etc.). The content script only observes the play, pause, and seek state of the page's primary <video> or <audio> element. It does not read other page content, form input, or cookies.

Data handled by the extension is used solely to provide the live-commentary feature. Specifically, we do not:

• Sell or rent any user data.
• Use captured audio, transcripts, URLs, or any other data for advertising or to build user profiles.
• Transfer data to third parties for purposes unrelated to generating commentary, except as required by law.
• Allow humans to read transcripts except where strictly necessary to fix a bug, respond to a user request, or comply with a legal obligation.

Raw tab audio is never written to disk on our servers. It is streamed through real-time speech-to-text and discarded. Anonymous session metadata and transcripts (text only) may be retained indefinitely to help us debug issues, prevent abuse, and improve quality. Because no record is linked to a user identity, there is nothing personally identifying to delete on your behalf, but we will delete the records associated with a specific session ID on request. See contact below.

Couchverse is not directed at children under 13 and we do not knowingly collect data from them.

If we materially change how data is handled, we will update the “Last updated” date above and, where appropriate, surface a notice in the extension.

Questions or deletion requests: eos@lemonslice.com.