Snapshot 310310
Normalized text
Scripts and page chrome removed; this is what change detection compares.
Skip to content LEGAL Privacy Policy How VoxisLive collects, uses and protects your information. Effective date: June 17, 2026 · Operator: VoxisLive Limited (England and Wales, company no. 17500296, ICO registration no. ZC271504) · [email protected] · Last updated: November 19, 2026 1. What we collect Account information — email, display name and a hashed password, stored in our account system for authentication. Signing in with Google or Apple — if you choose to sign in with Google or with Apple, that company confirms who you are and gives us your email address and an identifier for your account with them. We do not receive your password, and we do not ask for or keep your name. With Apple you can choose to hide your email address; Apple then gives us a private relay address that forwards our messages to you, and we never see your real address. To a relay address we send only what your account needs (sign-in and security mail, receipts and account messages) and, once, a personal note from our founder; offers and newsletters go to it only if you opted in to them. Apple forwards mail only from senders we have registered with it, so the receipt from our payment provider (Polar) cannot reach a relay address: the checkout page asks you for an email address for your receipt instead. Google and Apple act as independent controllers for their sign-in services, under their own privacy policies. Billing data — a customer ID from our payment processor. Card details are handled entirely by the processor; we never see or store your full card number. Usage metrics and diagnostics — translation minute counts and session metadata, used to enforce plan limits, and the technical outcome of each session (for example connection errors, delays and audio levels; never the audio or the text), used to find and fix sessions that fail. Device data & Anti-Abuse — a one-way, non-reversible cryptographic hash of a hardware identifier (device_hash) to prevent free-tier trial abuse; the raw identifier is never retained or uploaded. Live audio streams — captured audio is transmitted transiently over encrypted WebSockets (wss://) to Google Cloud (Gemini Live API) and Alibaba Cloud (Qwen Realtime API) for real-time translation. We do not store or retain your audio on our servers. Under enterprise API terms, transient audio is not used to train AI models. The app shows you this once, before your first session, and asks you to confirm it before any audio is sent. Live caption sharing (optional, Meeting mode) — if you turn it on, the translated captions of your meeting pass through our server, in memory only, to the people who have your link; if you offer extra languages, the caption text is also sent to Google (Gemini API) to be translated. Before your first share we record your confirmation of the sharing notice. See section 12. On-device speaker-change detection — speaker-change labels (S1/S2) are computed 100% locally on your computer CPU using embedded sherpa-onnx models, and the voice embeddings never leave your device. This marks when the voice changes; it does not identify who is speaking, cannot assign a name, and loses accuracy over overlapping speech or heavy music. Website assistant — if you use the chat on voxislive.com, the questions you type, the answers, the page and language, and the country of the connection (no IP address) are stored on our server for 30 days, with email addresses, keys and card-like numbers masked. The question and the conversation so far are sent to Anthropic (Claude API) to write and check the answer. Offer waitlists — if you leave your email address on the site to hear about an offer (for example Black Friday early access) and future Voxis offers, we store the address, your language, the page you signed up on, and when you signed up and confirmed — no IP address. If the same address has a Voxis account, the entry is linked to it. We first confirm the address by email (double opt-in); the confirmation and offer emails are sent through Resend, every message carries an unsubscribe link, and unconfirmed addresses are deleted after 30 days. Emails you send us — kept in our mailboxes (Google Workspace). To prepare our reply, the text of your message and a short summary of your account (plan, minutes, recent session outcomes) are sent to Anthropic (Claude API). Routine replies can be sent without a person reading them first; anything about a refund, a discount, a promise or an unverified fix waits for a person. WhatsApp support messages — if you are a paying member and message our WhatsApp support line, we receive your phone number (or, if you use a WhatsApp username, the WhatsApp user ID that Meta assigns for our business — in that case we may not receive your number), your WhatsApp profile name and what you send us there (text, images, voice notes). These messages are read and answered only by people on our team; they are not sent to an AI service. Local storage & Security — your browser stores a session token and language preferences; the desktop app stores settings, and any locally-held legacy BYOK API key is encrypted using Windows Data Protection API (DPAPI). Without an email address — with a password, or through Google or Apple sign-in — we cannot create an account, and without payment details we cannot process a purchase. Nothing else on this list is required to use the service. 2. How we use it To operate the real-time translation platform and your account, process billing through third-party payment systems, enforce usage quotas and prevent free-tier misuse, find and fix sessions that fail, answer your emails, and send transactional email via Resend. If you are a paying member and message our WhatsApp support line, we use your phone number or WhatsApp user ID to link the chat to your Voxis account and to answer you there; replies are written by people on our team, not by an AI. We also keep a pseudonymous record of each support case to check and improve the quality of our support (see "Retention and deletion"). Legal basis. For each purpose we rely on one of the grounds the UK and EU GDPR allow: Legal basis by purpose Purpose Legal basis Account, sign-in, licence and minute balance Our contract with you Live translation of the audio you choose Our contract with you Speech of other people that is part of that audio Our legitimate interest in delivering the translation you asked for; it is processed only while the session runs and is not stored Live caption sharing: relaying the captions you share to your listeners, and translating them into the extra languages you offer Our contract with you Live caption sharing: listeners' connection data, chosen language, waiting number and ticket; other people's words that are part of the shared captions Our legitimate interest in delivering the sharing you asked for and in protecting the service; processed only while sharing is on, and the captions are not stored Record of your confirmation of the sharing notice Our legitimate interest in showing what was agreed and in establishing or defending legal claims Usage metering and plan limits Our contract with you Payments and receipts Our contract with you; keeping the records is a legal obligation Support: emails, WhatsApp messages and problem reports Our contract with you; our legitimate interest in answering you if you are not a customer Improving our support: pseudonymous WhatsApp case records Our legitimate interest in checking and improving the quality and accuracy of our support Diagnostics Our legitimate interest in finding and fixing sessions that fail One free tier per device Our legitimate interest in preventing abuse of the free tier Security of the website and the API Our legitimate interest in protecting the service Website assistant Our legitimate interest in answering questions before you buy Marketing emails Your consent; if you have bought from us, our legitimate interest. Every message has a link to unsubscribe Accounting records Legal obligation 3. Third-party sub-processors Data flows only to the sub-processors needed to deliver the service: Polar — payments and invoicing, as merchant of record. Card details are handled by Polar; we receive a customer ID and the order record. Our own server (PocketBase) — account, licence and usage records, on a server in the United Kingdom that we rent from IONOS Cloud Ltd and administer ourselves. The same server relays shared captions (live caption sharing) without storing them. Resend — email delivery: sign-up, purchase receipts and support replies, and the confirmation and offer emails of the offer waitlist. Google Workspace — our mailboxes (support@, contact@ and the founder's address): the emails you send us and our replies. Meta (WhatsApp Business Platform) — our WhatsApp support line for paying members, used only if you message us there: your phone number or WhatsApp user ID, your WhatsApp profile name, the messages, images and voice notes you send us, and our replies. WhatsApp is operated by WhatsApp Ireland Limited / Meta Platforms Ireland Limited under Meta's business terms. Google Cloud / AI Studio — Gemini, live audio translation; translation of a video's caption text for the browser extension's synced dub (text only, no audio); and, only if you offer extra languages in live caption sharing, translation of the caption text into those languages. Alibaba Cloud / DashScope — Qwen, live audio translation. Cartesia — Sonic text-to-speech, for the desktop app's voice on the free daily minutes (and, where we have switched it on for an account, its natural voice in paid sessions) and for the browser extension's synced dub on videos with captions and its live translation on the free daily minutes: it receives only the text of the translated sentences (no audio, no account identifier, no email address, no video address). The app or 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. Cartesia AI, Inc., United States; we have switched off the use of our data for Cartesia's model training. Cloudflare — delivers and protects voxislive.com (content delivery, attack protection, and the Turnstile bot check on sign-up and the website assistant). Requests to our website and API, including your IP address, pass through Cloudflare's network; live translation audio does not. Requests to cn-api.voxislive.com, our direct route for mainland China, do not pass through Cloudflare; for those, our own server looks up the country of your IP address in a local database, and the address is not sent anywhere for this. IP geolocation by DB-IP. Cloudflare R2 also stores the offsite copy of our nightly backups, kept for 14 days. Anthropic — Claude API, for the website assistant (the chat on voxislive.com), where it receives the questions you type and the conversation so far, and for drafting our replies to your emails, where it receives your message and a short account summary. It is not used on our WhatsApp support line. Under Anthropic's commercial API terms, inputs and outputs are deleted within 30 days and are not used to train models. We never sell or rent your data. The full list, with what each one receives, is on the sub-processors page. 4. Browser extension privacy In Chrome, Edge and other Chromium browsers, the Voxis browser extension uses tabCapture permissions to capture active tab audio in volatile memory for live translation overlays. Tab audio is processed in RAM only while translation is active and is never stored, profiled, or shared with advertisers. The Safari extension, which comes inside the Voxis Mac app, has no such permission and does not capture tab audio. It dubs a video from its captions: after you click Start it reads only the video's playback time and its caption track, and the caption text is sent for translation and speech to the AI services listed in section 3. The Safari extension and the Voxis Mac app also exchange a few small files in a folder on your Mac that macOS shares between them: the app's status (its version, and whether it is running, translating and signed in), a request to open the app with your target language, and, only if you choose to continue with the account in the Voxis app, that account's sign-in token, which is deleted from the folder as soon as the extension has read it. These files stay on your Mac and are not sent to us. The details are in the Browser Extension Privacy Policy. 5. Retention and deletion What we keep, where it lives, and for how long: Retention by data type Data Where it lives How long Account data (email, display name, password hash, and the identifier of a linked Google or Apple sign-in) Our server While your account is active Usage events (minute counts, session metadata) Our server Kept with the account; deleted or anonymized when the account is deleted Problem report (scrubbed log, diagnostics, and the transcript only if you tick the box) Our server 12 months after the case was last updated; earlier on request Diagnostics (technical outcome of each session) Our server 13 months Device hash (free-tier protection) Our server 24 months after the device was last seen Emails you send us, and our replies Our mailboxes (Google Workspace) Up to 6 years, because messages about purchases, refunds and complaints are part of our accounting records; other messages are deleted on request WhatsApp support chat (phone number or WhatsApp user ID, messages and media you send, our replies) Our systems, and Meta's WhatsApp platform 30 days on our systems, then deleted automatically; the link between your phone number or WhatsApp user ID and your account stays while the account exists, or until you ask us to remove it. Meta keeps WhatsApp messages under its own terms. Pseudonymous WhatsApp case records (your messages and our reply with email addresses, phone numbers, user IDs and codes removed; how the case was classified and handled) Our systems Kept for as long as they help us check and improve our support; they carry no name, number or email, only a random case identifier. Ask us and we delete the records of your cases. Record of the marketing emails sent to you Our server 24 months Backups of our server Our server and Cloudflare R2 14 days Transcripts and recordings Your own disk — they are not uploaded to our servers. Live caption sharing, if you turn it on, relays captions through our server in memory and does not store them (section 12) Until you delete them AI session summary Sent to the model only when you press the button; the text comes back to your computer Not stored on our servers Live audio Passes through the translation engine while the session runs Not stored Website assistant conversation (masked text, page, language, country, rating) Our server 30 days, then deleted automatically Offer waitlist (email address, language, sign-up page, sign-up and confirmation time) Our server Unconfirmed: 30 days. Confirmed: until you unsubscribe; after that the entry stays marked as unsubscribed so we never email you again — ask us and we delete it Shared captions (live caption sharing) Our live-sharing server, in memory only Until sharing ends; not stored Share log of live caption sharing (opened/closed, languages, duration, counts; no captions, no account) Our server's system log Up to 30 days Record of your confirmation of the caption-sharing notice (account and licence identifiers, notice version and fingerprint, date and time, app version, language; no captions, no IP address, no email address) Our server While you have an account, and six years after it is deleted How to delete. Transcripts you delete yourself, from the saved-transcript list in the app; the file leaves your disk. For the account, write to [email protected]; we will delete or anonymize your account data within 30 days, except where legal or accounting obligations require retention. You can also delete the account yourself in the app: Settings › Membership › Delete my account. One record stays after deletion: if you confirmed the live caption sharing notice, the record of that confirmation is kept for six years, without your email address (see section 12). If you signed in with Apple, deleting your Voxis account also removes the link to your Apple ID on our side; you can additionally stop using Sign in with Apple for Voxis in your Apple Account settings, which on its own does not delete your Voxis account. 6. Your rights Depending on your jurisdiction (EU GDPR, UK GDPR, California CCPA, Türkiye KVKK), you have the right to access, correct, export, delete, restrict or object to the processing of your personal data. Where we rely on your consent, you can withdraw it at any time; this does not affect processing that already took place. You can object to processing based on our legitimate interest, and you can stop marketing emails at any time with the unsubscribe link or by writing to us. We answer requests within one month. We do not make decisions about you by automated means that have legal or similarly significant effects, and we do not build profiles of you for advertising. If you are not satisfied with our answer, you can complain to a data protection authority. In the UK this is the Information Commissioner's Office (ICO): ico.org.uk, telephone 0303 123 1113. In the EU it is the authority in the country where you live. 7. Security Passwords are hashed with argon2/bcrypt; Legacy BYOK API keys are encrypted with hardware-bound Windows DPAPI; all transmission uses TLS/SSL (wss://, https://). On-device processing covers speech detection (VAD) and speaker-change labelling. The free tier's voice is produced either on your device or by a cloud voice service (Cartesia); when the cloud voice is used, only the text of the translated sentences is sent to it, under the same terms as the rest of the translation. On paid plans the captured audio streams continuously to the translation model while a session is live; on the free tier the stream is gated, so only detected speech is sent. 8. Children The service is not directed at and may not be used by anyone under 18. 9. Contact & Data requests Privacy questions and data protection requests: [email protected] or the contact page. 10. Cookies and browser storage The website sets one cookie, voxis-lang, which remembers your language for up to a year. The account pages keep your sign-in in your browser's own storage, and the site uses that storage for a few things it needs to work: your language, the website assistant conversation and a purchase you have started but not finished. All of this stays in your browser. We use no advertising or tracking cookies, and our page counts work without cookies. Because everything here is needed for the site to work, we do not ask for cookie consent. The live caption page (voxis.live) sets no cookies; it keeps your text-size choice in your browser's local storage, and the language you chose, which side of the meeting you show and a random reconnect ticket in session storage, which your browser drops when you close the tab. 11. Changes to this policy When we change this policy, we update the date at the top of the page. If a change affects how we use your data in a significant way, we also tell account holders by email. 12. Live caption sharing (optional) Live caption sharing lets other people read, on their own phones or computers, the live translated captions of your meeting. It is off unless you turn it on, and it works only in Meeting mode of the desktop app. It is available only where we have switched it on for your account: it is offered on paid accounts (a plan or a minute pack), on promotional accounts that we grant and on seats under an organisation plan, and not with your own API key. If your seat is provided by an organisation, that organisation is responsible for how its seats are used and for what its members share, and we handle the shared captions on its behalf under our data processing agreement with it; we remain responsible for the confirmation record described below and for the technical data of listeners. Your organisation may have its own rules about sharing a meeting. What is shared. A share carries the translation of what you say, in the language you are speaking to the other side in. A share can also carry the translation of what the other people in your meeting say: the app asks you to confirm this each time you start a share, and without your confirmation the other side's words are not shared. The app sends the translated caption lines, as they are produced, to our live-sharing server, which passes them on to your listeners. A line may be updated or withdrawn while it is still being translated. With each line we pass the language of the captions and, when both sides are shared, which side it comes from. A line from the other side can carry a speaker number (S1, S2); it marks that the voice changed and does not say who is speaking. What is not shared. Audio is not shared: neither the original sound nor the translated voice. The original-language text is not shared. Your name, email address and other account details are not shown or sent to listeners. Transcripts and recordings saved on your device are not uploaded. The translation of what the other side says is not sent to the live-sharing server unless you confirmed it for that share. Limits you should know about. "What you say" means what your microphone hears. If other people speak near your microphone, or if sound from your speakers reaches it, their words can be translated and shared as if they were yours. Using headphones makes this much less likely. The app can pause the sharing of your side by itself when it detects sound from your speakers in your microphone; it then tells you, lines that overlap the pause are not sent to listeners in any language, and you can resume when you are ready. This is a safeguard, not a guarantee: it will not catch every case, and it does not detect another person speaking in the room. The other side's captions are made from the sound your computer plays during the meeting, so other speech playing on your computer at that time, such as a video or a second call, can be captioned and shared as well. Captions in a listener's own language. You can offer your listeners up to 5 more languages. Each listener then chooses, on the caption page, which of the offered languages to read. To do this, the app on your device sends the text of the captions you share (completed sentences only) to an AI text-translation service (Google, Gemini API), receives the translations and sends them to our live-sharing server, which passes them on. When a caption line of several sentences is finished, the app sends that line once more, together with the translations already shown, so that the service can correct them in the light of the whole line; a listener may therefore see a sentence change shortly after it appeared. A language is translated only while at least one listener is reading in it; if you offer no extra languages, nothing is sent to that service for this feature. Only caption text is sent: no audio, and nothing about your listeners. These captions are a translation of a translation, so they can contain more mistakes, and they appear sentence by sentence rather than word by word. Google acts as our processor (see section 3). Under Google's terms for the paid Gemini API, Google does not use this text or the translations to train its models or improve its products. Google keeps a log of the requests and responses for 55 days, only to detect and prevent abuse of its service; for that purpose authorised Google staff may look at them. Google may process this data in any country where it or its agents have facilities, so it can leave the United Kingdom. Where the captions go. The captions travel over an encrypted connection from your app to our live-sharing server and from there to your listeners' browsers. For each language the server keeps up to the last 30 completed lines in memory, and the lines currently being translated, so that a listener whose connection drops can pick up where they left off. A person who joins later does not receive what was said before they joined: they see the sentence being spoken at that moment and what follows. The live-sharing service does not write captions to disk or to a database, so they are not in our backups. When sharing ends, the captions in every language are removed from the server's memory. Sharing ends when you stop it, when you stop the meeting, when you close the app or sign out, when the app has not contacted the server for about 10 minutes, or after 8 hours at the latest. Our live-sharing server is located in the United Kingdom, at our hosting provider (see section 3). Who can read them. Anyone who has the link, the QR code or the 8-character code can read the captions while sharing is on. You can copy the link or save the QR code as an image; anyone you give it to, and anyone they pass it on to, can use it for as long as sharing is on. Listeners do not sign in and we do not know who they are. Up to 100 people can read at the same time. Please treat the link like a public link: listeners can copy the text or take screenshots, and we cannot control what they do with it. Asking before someone joins. You can switch on "ask me first" (it is off by default). A person who opens your link then waits, and receives no captions in any language, until you let them in. You see a random two-digit number for each waiting person and how long they have waited, and nothing else. The number does not tell you who the person is and is not a check of their identity. If you are a listener. You do not need an account and we do not ask for your name or email address. You see the captions from the moment you join, not what was said earlier. The captions are published by the person sharing them; we do not review them. When that person ends sharing, the captions are removed from the page in your browser; anything you copied or captured yourself is outside our control. The listener page sets no cookies and loads no analytics or advertising scripts. A few small things are kept in your own browser: your text-size choice (local storage; it is not sent to us), and, in session storage, the language you chose to read in, your choice of which side of the meeting to show, and a random ticket that lets the page reconnect to the same share; your browser drops these three when you close the tab. The page suggests a language from your browser's language settings; the language you read in is sent to our server so that it can send you the right captions, and the speaker's app is told how many listeners are reading in each language (a count, not who they are). Our server keeps only a one-way fingerprint of your ticket, in memory. If the speaker has chosen to approve listeners, the page shows you a random two-digit number; the speaker sees that number and how long you have waited, and can let you in or decline. The speaker does not see your IP address, your device or your location. The ticket, the number, the decision and your language choice are held in memory only and are deleted when sharing ends. To deliver the captions and to protect the service from abuse, we process your IP address: it is used to keep the connection open and to apply limits (how fast and how many connections come from one address, repeated wrong codes, and repeated declined requests). The address is held in the server's memory only: for the limits, for a short time (minutes); together with your ticket, until sharing ends. The live-sharing service does not write it to its logs. Nothing about you is sent to the translation service. If you are the one sharing. We check your sign-in and your plan to confirm that your account may use the feature, and keep your licence identifier in the server's memory while your share is open and for a short time afterwards, to apply limits. The service's log records that a share was opened and closed, with its caption languages, which sides were shared, whether "ask me first" was on, how many extra languages were offered, its duration, the highest number of listeners, how many caption updates were sent, and how many join requests were accepted, declined or timed out. These log entries contain no caption text, no codes, no tickets, no listener numbers and no IP addresses, and they do not name your account. Separately, your app reports that the feature was used as part of the diagnostics described in section 1: for example whether a share was paused, whether the other side was included, and listener and request counts in ranges. These diagnostics are kept with your account like the others and never contain the captions. Record of your confirmation. Before your first share, and again whenever the in-app notice changes, the app shows you the notice and asks you to tick a box confirming that you are responsible for what you share. Sharing does not start until that confirmation has been recorded on our server. We keep a record of it: your account and licence identifiers, the version of the notice, a fingerprint of the notice text in the language it was shown in, the version of the Terms in force, the date and time, the app version, the distribution channel and the app language. The record contains no captions, no share code, no IP address, no device identifier and not your email address. We keep it for as long as you have an account and for six years after the account is deleted, so that we can show what was agreed if a dispute or legal claim arises; six years is the period within which such claims can normally be brought in England and Wales. After that it is destroyed at our next routine clean-up, within six months. This record is kept even if you delete your account. When you delete your account, we add to the record a one-way, keyed fingerprint of your email address and the date of deletion; the fingerprint lets us match the record to you only if you, or a claim concerning you, give us that address. If a claim or complaint about a share is made during that time, we keep the record until the matter is finished. The box you tick at the start of a share to include the other side's words is not recorded on our server. Other people. What you say can be about other people, your microphone can pick up other people's voices, and, where the other side is shared, the captions contain what the other people in your meeting say. If you offer extra languages, those words are also sent to the translation service as part of the caption text. You decide whether to share and with whom, and you are responsible for telling the people in your meeting and for getting their permission where it is needed; see the Terms of Service, section "Live caption sharing". Legal basis. For the person who shares: our contract with you, because you asked for this feature, including the extra languages you choose to offer. For listeners, and for other people whose words end up in the captions: our legitimate interest in delivering the sharing that the sharing user asked for, making speech understandable across languages, and keeping the service secure; the processing is brief and we keep no captions afterwards. For the record of your confirmation: our legitimate interest in being able to show what was agreed and to establish or defend legal claims. How long we keep it. Data Where it lives How long Shared caption lines, in each offered language Our live-sharing server, in memory only Up to the last 30 completed lines per language, until sharing ends Caption text sent to the AI text-translation service (only if you offer extra languages) Google (Gemini API) Not kept by us. Google logs requests and responses for 55 days for abuse detection only and does not use them to train its models Listener ticket (one-way fingerprint), waiting number, the speaker's decision, the listener's chosen language Our live-sharing server, in memory only Until sharing ends Listener IP address Our live-sharing server, in memory only For rate limits, minutes; with the ticket, until sharing ends. Not written to the live-sharing service's logs Share log (opened/closed, languages, sides, duration, counts; no captions, no account) Our server's system log Up to 30 days Record of your confirmation of the sharing notice Our server While you have an account, and six years after it is deleted (longer only while a claim is open) Your rights. The rights described in section 6 apply. Because we do not keep the captions, and keep nothing that identifies a listener after sharing ends, we will usually have nothing to find, correct or delete afterwards, and we may not be able to identify you in this data at all. The record of your confirmation is different: you can ask us for a copy of it, and you can object to our keeping it; we will then stop unless we need it for legal claims, and for the same reason we may decline a request to erase it before the six years have passed. You can contact us at [email protected], and you can complain to a data protection authority as described in section 6. To report a shared link that is being misused, write to [email protected]. Keep reading Extension PrivacyHow the VoxisLive browser extension handles your data: tab audio is streamed for translation, never retained, with…TermsThe agreement between you and VoxisLive when you create an account or use the service.AboutVoxisLive delivers live spoken translation of any audio on Windows, Mac and Linux — driverless, 79 languages.