What the extension does
The VoxisLive browser extension captures audio from your active browser tab and streams it for translation to Google's Gemini Live API, or to Alibaba Cloud's Qwen real-time API for the languages it voices. It returns spoken translations along with captions that are displayed locally on your device.
On the free daily minutes, Google (Gemini) translates the tab audio into text (on a video with captions, the synced dub described below runs instead and uses the caption track), and the translated sentences are then spoken aloud by Cartesia, a text-to-speech provider, or by your browser's own voice when Cartesia is unavailable or does not speak your language. Cartesia receives only the translated sentences — never the tab audio, the page or its address. Your browser connects to Cartesia directly with a short-lived token our server issues, so Cartesia technically sees your IP address; no account information is sent to it.
Audio is captured only while a translation session is running. When no session is active, the extension is not listening to your tab.
What data we collect
During an active session, the extension collects the audio of your current tab and streams it for translation. Recordings are not retained on our servers once the translation has been delivered.
In account mode, your email and password are transmitted to voxislive.com so you can sign in. Bring-your-own-key (BYOK) mode also requires a VoxisLive account, because running on your own key is covered by a purchased BYOK licence: when you open the extension and when a session starts, it asks voxislive.com whether that licence is valid, sending your account token and nothing else. It reports no session, no minutes, no languages and nothing about what is being translated, and no usage is recorded for own-key sessions. Your API key itself stays on your device and never reaches VoxisLive servers.
Preferences such as your target language, voice choice, volume, and authentication mode are stored only on your device.
Synced dub on videos with captions
When you start a session on a video that already has captions (YouTube and many other video sites), the extension can dub from those captions instead of from the audio, so each translated sentence is spoken exactly when the speaker says it. After you click Start, a small script runs in that tab only. It reads the video's playback time and the caption track the page has already loaded; to get that track it may switch the player's captions on in the background for a moment, then puts your caption setting back. It does not read forms, passwords or anything else on the page, and nothing runs before you click Start.
The caption text is sent to Google (Gemini) for translation, and the translated sentences are turned into speech by Cartesia, by Google, by Alibaba Cloud (Qwen) or by your browser's own voice. Cartesia receives only the translated sentences, never the original captions, the video or its address. Your browser connects to Cartesia directly, so Cartesia technically sees your IP address; no account information is sent to it. The caption text and its translation go directly from your browser to these providers; voxislive.com never receives them. The tab audio is still captured during a synced dub, but it stays on your device: it is used only to lower the original voice and to count session minutes, and it is not sent anywhere. If a video has no usable captions, the session falls back to the live translation described above.
To decide whether a synced dub is possible, the extension checks the tab's address against a fixed list of video sites on your device. Only a short site label (for example youtube) is sent to voxislive.com — never the address, the video, its title or its captions.
If the video plays inside a player embedded from another site (YouTube, Vimeo, Dailymotion, Wistia or Brightcove), the extension asks your permission once to run the same caption script inside that player. This permission is not granted at install, and if you decline, live translation keeps working.
Who your data is shared with
Google (Gemini) or Alibaba Cloud (Qwen) receives the tab audio in order to perform the translation, depending on the target language and, in BYOK mode, on which key you supplied. This is inherent to how the real-time translation service works.
In account mode, voxislive.com receives your sign-in credentials, session keys, plan quota, and usage metrics — but never your audio. In BYOK mode, voxislive.com receives one thing: the licence check described above, sent when you open the extension and when a session starts. Your audio and your API key never reach VoxisLive in either mode. A check that has just succeeded is reused for a short period rather than repeated, and a licence already proven on this device keeps working for a limited time without a connection, so an own-key session does not depend on us being reachable.
For a synced dub, Google receives the video's caption text instead of the tab audio, and Cartesia, Google or Alibaba Cloud turns the translated sentences into speech; Cartesia receives only those translated sentences. Your browser connects to Cartesia directly, so Cartesia technically sees your IP address; no account information is sent to it. voxislive.com receives only the site label described above.
On the free daily minutes, Google receives the tab audio to translate it into text, and Cartesia — or your browser's own voice — speaks the translated sentences; Cartesia receives only those translated sentences, never the audio. Your browser connects to Cartesia directly, so Cartesia technically sees your IP address; no account information is sent to it.
What we deliberately do not collect
The extension does not track your browsing history, clicks, keystrokes, mouse movement, or scrolling. It does not collect location, health, or financial data.
Our scope is intentionally narrow: translate the tab audio you ask us to translate, and nothing more.
A synced dub reads only the playback time and the caption track of the video you started it on — nothing else on the page.
Retention and deletion
Audio and transcripts are not retained. Account holders can request deletion of their account data through our contact page, and requests are processed within 30 days, except where retention is legally required.
Security and your rights
Passwords are hashed server-side. Under applicable laws such as KVKK and GDPR, you may exercise your rights of access, correction, and deletion.
To make a privacy request, contact us at [email protected].