VoiceNear Privacy Policy

Last updated: 4 November 2026

1. Who we are

VoiceNear is made by Sangei AI Limited (神機智能有限公司), a company incorporated in Hong Kong ("Sangei AI", "we"). This policy explains what happens to your information when you use:

  • the VoiceNear apps for Mac, iPhone, iPad and Apple Watch;
  • the online services that support them;
  • the website voicenear.com.

Questions and requests: support@voicenear.com.

2. What stays on your device

Meetings are recorded on your device. In Local mode they are also processed there. In Balanced mode the speech is recognised on your device, and only the text described in section 3 is sent. In Online mode, if you have agreed, the meeting’s audio is sent while you record, as section 3 describes. The following stay on your device unless you export them or use a feature described in section 3:

  • your recordings, transcripts and translations;
  • recaps and reports made on your device;
  • saved voices, described below;
  • participant lists and the photos you take for them.

Saved voices. Recognising speakers by a saved voice is off by default in each meeting. To turn it on for a meeting, the organiser must confirm in the app that everyone who speaks has agreed, or the participants agree themselves. Participants can also agree after a meeting, for that meeting. A voice is saved only when the speaker gives their own agreement on the screen. VoiceNear keeps a dated record of that agreement, and the saved voice expires 2 years after the agreement. Saved voices stay on the device where they were saved (Mac, iPhone or iPad). They are matched there, and are never synced or uploaded. When you continue a meeting, VoiceNear re-reads the voices from that meeting's own recording, in memory only, and stores no voiceprint. The web app is different: it keeps no saved voices, but to number the speakers of a meeting, our service makes a voiceprint of each speaker from the audio you send and matches them for that meeting only, as section 3 describes. You can delete a meeting or a saved voice in the app at any time.

One device, several accounts. Each VoiceNear account sees only its own meetings on a device. Signing out hides them, and signing in with another account shows only that account's meetings.

Calendars. If you link your device calendar (the system asks you first), VoiceNear reads your events' titles, times, places and attendee names on your device, to show your meetings, and adds or changes an event only when you use the app's own calendar form. An event's notes are read only to find a video link and to show you the event; they are not stored. If you connect Google Calendar or Outlook, VoiceNear reads your events directly from Google or Microsoft, with read-only access, and reads the attachments of an event only to list them and to add the files you choose as documents in that meeting. Calendar data and those files stay on your device and are not sent to our servers. The one exception: when you start a meeting from an event, its name and its attendees' names can become the meeting's title and speaker names, and then go where section 3 says titles and names go.

Google Drive. Instead of iCloud, you can choose Google Drive to sync your meetings between your own devices. VoiceNear then saves each meeting's transcript, reports, notes and speaker names as a file in a folder of your own Google Drive. It has Google's "drive.file" access only, which reaches the files VoiceNear itself created or that you gave it, and nothing else in your Drive. Recordings, screen video, note pictures and handwriting are not sent. The files go between your device and Google, not through Sangei AI, which cannot read them; Google runs Drive under its own terms.

On the VoiceNear Local plan, which has no online features and no sync, nothing about your meetings goes to Sangei AI. On a plan with online services, Local mode sends nothing to Sangei AI unless you choose an online feature described in section 3. Re-read on my Mac, iCloud sync and Google Drive sync (sections 2 and 3) move your information only between your own devices and your own accounts. The app still downloads the speech and translation models it needs (section 4).

3. What leaves your device, and only when you use these features

Online translation.

  • In Online mode, each finished line's text, and the languages you chose, go to Sangei AI's service, which translates it. In Balanced mode, only the lines your device cannot translate itself go. In Local mode, none do.
  • A question you type in "Ask in their language" on the Mac is sent the same way.
  • When the app translates a whole meeting at once (Improve Translation on the Mac, or an imported recording on iPhone), each line goes with the meeting's title, the speakers' names and nearby lines, so that the translation fits the conversation. If the meeting started from a calendar event, that event's name and attendees' names can serve as the title and speaker names.
  • We do not store the text of live translation. We keep counts of lines and usage for your allowance and billing, not the content. A whole-meeting translation is the exception: it is held as "After a meeting" below says.
  • Translation itself uses text only. Audio is covered under Live speech recognition and Re-read and imports below.

Live recaps. On a plan with online services, in Online and Balanced modes, while the live recap is open, the app sends the text of the most recent lines, with speaker names, to Sangei AI's service, which writes the recap. We do not store the text. In Local mode the recap is written on your device, where the device can, and nothing is sent.

Live speech recognition. This is off until you agree. In Online mode, if you have agreed to your meeting audio being heard online, the app sends short pieces of the meeting’s audio (up to about 12 seconds each) to Sangei AI’s service while you record. The service turns the speech into text, translates it, and sends both back to your device. This works on iPhone, iPad and Mac. A Watch that records on its own sends its pieces the same way, and its finished recording then goes to your paired iPhone (see below). Your recording is still saved on your device. In Local and Balanced modes no audio is sent while you record. You agree with one switch in the app’s settings; turning it off stops all audio being sent, live and Re-read, and recording carries on. We do not store the audio. We keep its text only until the translation and report are done, 24 hours at most, as "After a meeting" below says. We keep counts and timings for your allowance and billing and, for 72 hours, a one-way fingerprint of each piece so that a repeated send is not counted twice, never the audio or the text.

Re-read and imports. A saved recording is uploaded to our service only in two cases: you tap Re-read on a saved meeting (on iPhone, iPad or Mac), or you choose Online when you import a recording. Both need the same agreement as live speech recognition. It is transcribed by Sangei AI's service. We do not keep the audio. We keep the new transcript only until the translation and report are done, 24 hours at most, as "After a meeting" below says. If you choose "On this iPhone only" when importing, nothing about that recording leaves your device.

Cloud reports.

  • When the online service writes a report for you (the default in Balanced and Online modes), the transcript is sent to our service. Sangei AI's service writes the report.
  • The job, including the transcript and the result, is deleted from our service as soon as your device collects it, and at the latest 24 hours after it started.
  • A whole-meeting translation job (its text and its result) is held the same way.

After a meeting. When a meeting that used Online mode ends, the app can send the meeting's corrected transcript once, so that the report and the whole-meeting translation are made from it. Our service keeps that transcript only until the report and the translation are done and deletes it at once after; whatever happens, never longer than 24 hours.

Speaker numbers and voiceprints. On the web app, when you send audio online (live or an import), our service makes a voiceprint of each speaker from each piece of audio and matches it with the other voices heard in the same meeting, so that each voice has the same number (Speaker 1, 2 and so on) through the meeting. For that meeting only, the service holds one averaged voiceprint per speaker and the numbers it has given: no audio, no name. It does not write them to our database or our logs, uses them for nothing else, and deletes them 15 minutes after the meeting's last piece of audio, and at most 12 hours after its first. On the Mac, iPhone and iPad, voiceprints are matched on the device (section 2); in Online mode our service still numbers the speakers it hears for the live lines, in a temporary record kept and deleted by the same rules.

Questions about your meetings. When you ask a question about a meeting, your question, the meeting's transcript, your saved Memory, and your notes and attached document pages from that meeting may be sent online to be answered, and the answer comes back to you. Sangei AI does not keep them after the answer is sent. The recording is not sent. Slide pictures are not sent. A meeting recorded in Local mode is never sent. Your questions and answers are kept in your history on your devices and, if you turn on iCloud sync, in your iCloud.

When you ask about a folder or about all your meetings, only the matching passages are sent, each labelled with its meeting's title and date, plus your Memory and notes and document pages that may relate to the question. Folder names are not sent. Meetings recorded in Local mode are never searched for these questions.

Names and participant lists. On a cloud plan, the app can suggest a meeting title and speaker names, and read a participant list or business card from a photo. For these, the transcript lines and participant list they need are sent to Sangei AI's service. The app reads a photo on your device and sends only the text it finds. Only use photos of lists or cards you are entitled to share.

Feedback. When you send feedback from the app, the message comes to us, with the topic you chose. A diagnostic log is sent only if you choose to attach one. We keep a reply email address only if you tick that we may reply to you and type an address. A one-tap rating after a meeting sends the rating with that meeting's mode and its length in minutes. Feedback never includes meeting text or audio.

Your own Mac and your own iCloud. Two features move your information only between things that are yours, and none of it passes through Sangei AI:

  • "Re-read on my Mac" sends a meeting's recording from your iPhone to your paired Mac, over your own network.
  • A meeting recorded on your Watch goes to your paired iPhone.
  • If you turn on iCloud sync, your transcripts, reports and speaker names sync through your iCloud account, which Apple runs under its own terms. Your recordings sync too, by default, and a setting turns that off. Sangei AI cannot read anything you sync, and each VoiceNear account's meetings are kept apart in your iCloud. Sync starts off on each device for each VoiceNear account, and meetings reach another device only when it uses the same VoiceNear account and the same Apple ID.

4. Downloads

The app downloads its speech and translation models from public model-hosting services or our own download server; like any download, the server sees your IP address.

5. Accounts and purchases

When you create an account or buy a plan, we keep:

  • your email address, which may be the private-relay address Apple gives us if you use Sign in with Apple (with email sign-in, we confirm the address by sending you a link);
  • the identifier from Sign in with Apple or Google, if you use one;
  • your plan, subscription status and online allowance;
  • for each device you activate: its name, its platform and an installation identifier. A licence allows three devices, except the VoiceNear Local plan, which allows five places; each place can move to a new device once every 12 months, and we keep a record of each move;
  • the version and time of the online processing agreement you gave on each device, with no text or audio, so that we can show you what you agreed to and when, and know which wording applies.

Payments. Purchases on our website are handled by our payment processor: we receive the payment status, not your card number. Purchases in the iPhone or iPad app are handled by Apple, which takes the payment. We receive the purchase record tied to your account (the transaction, the product, its dates, and whether it was renewed, refunded or ended), check on our server that Apple signed it, and receive Apple's notices about renewals and refunds. We do not receive your card or Apple payment details.

Free trial. A free trial starts at your first sign-in, once your email address is confirmed, once per account and once per device. To enforce that, we keep the trial's dates and allowance and a one-way code made from the device's installation identifier; the code cannot be turned back into the identifier.

Sign-in links are sent by email.

Country of each purchase. For each plan you buy, we record separate country facts as short country codes only, never your address: the country you chose at checkout, the billing, tax and card countries as our payment processor reports them, the country our network reports for your connection at checkout, and, for App Store purchases, your App Store storefront. We use them to apply each country's consumer and tax rules, and to prevent fraud. We keep them while your account exists and delete them when you delete your account.

6. The website

voicenear.com uses no cookies or advertising trackers. The sign-in pages on api.voicenear.com, where you redeem a beta code or buy a plan, set only the cookies sign-in needs, and no trackers. To deliver pages and protect the site, our hosting provider processes technical information such as your IP address and browser type, and keeps it in its logs for a limited period. The website counts visits without cookies and without building a profile of you. The apps do not use this.

7. The beta list

If you join the beta list, we keep:

  • your email address;
  • the language of the page you signed up from;
  • the link you came from, if it had a tag;
  • the time you gave consent.

We use it only to invite you to the VoiceNear beta and to send you beta news, including one short email, about a week after you start using the beta, asking for your feedback. We do not sell it or share it for anyone else's marketing. An invitation carries a personal code that works once.

You can leave the list at any time with the form on the download page, the link in any beta email, or by writing to support@voicenear.com. Leaving also cancels an invitation code you have not used. We delete the list within 30 days after the beta ends.

8. Who processes information for us

We use providers in these categories:

  • hosting and infrastructure (our service, storage, email routing);
  • AI providers (translation, recaps, reports, answers to questions about your meetings, titles and name suggestions, reading participant lists);
  • speech recognition (live meetings you agree to send, Re-read and imports);
  • email delivery (sign-in emails and beta invitations);
  • payment processing (our payment processor on the website, Apple's App Store in the apps);
  • sign-in (Apple or Google, when you choose them);
  • public model-hosting services (model downloads; they see your IP address, as section 4 says).

Each processes information to provide its part of the service, under its own terms and security commitments. Our service keeps the text and audio it passes to AI providers and speech-recognition providers only as section 10 describes. Those providers may keep what they receive, under their own terms:

  • some keep it for up to 30 days, or longer only if the law requires it or the security of their service needs it, and for up to 2 years if their automatic safety checks flag it; their terms do not let them use it to train their models without separate consent;
  • others keep it for as long as their own terms allow, which may set no fixed period, and their terms may let them use it to improve their services.

Your own calendar and Drive accounts. If you connect Google Calendar, Outlook or Google Drive (section 2), your device talks to Google or Microsoft directly. They do not process it for us; their terms with you apply, and our servers receive only what section 3 lists.

9. International transfers

Our providers operate servers outside Hong Kong. Information sent to them may be processed in other countries or regions. We choose providers that protect the information they process and send them only what the feature needs.

10. How long we keep things

WhatHow long
Live translation text onlineNot stored. A whole-meeting translation is held until your device collects it; 24 hours at most
Audio and transcript for live speech recognition, Re-read and importsWe do not keep the audio. Its text is kept only until the translation and report are done, 24 hours at most. A one-way fingerprint of each live piece (no audio or text), used only so a repeated send is not counted twice, is deleted after 72 hours.
Online report job, and the corrected transcript after a meetingUntil the report and the translation are done and your device collects them; 24 hours at most, whatever happens
Speaker voiceprints and numbers (web app; speaker numbers on the apps)15 minutes after the meeting's last piece of audio; 12 hours at most
Account, devices, plan, online processing agreement record, purchase recordWhile your account exists; deleted within 30 days after you ask us to delete it, except billing records we must keep by law
Country facts of a purchase (country codes only)While your account exists; deleted with your account
Beta list and invitation codesUntil you leave the list, or 30 days after the beta ends. We then delete your address. A hash of each code (never the code itself) stays, attached to no address, and is no longer linked to the account that redeemed it once that account is deleted.
Free-trial recordThe trial's dates and the one-way device code stay after your account is deleted, with no link to your account or address, so that a device cannot start a second trial
Feedback you send from the appKept until we have handled it. A diagnostic log you choose to attach is deleted after 30 days. A reply address, if you gave one, is deleted with the feedback, and in any case when you delete your account. Feedback never includes meeting text or audio.

11. Your rights

You can ask us to:

  • give you a copy of the personal data we hold about you;
  • correct it;
  • delete your account and its data;
  • stop using it for the beta list.

Write to support@voicenear.com from the email address on your account or list. We will answer within 40 days, as Hong Kong law requires.

If you are unhappy with our answer, you can complain to Hong Kong's Privacy Commissioner for Personal Data, or to the data protection authority where you live.

Deleting your account in the app. Before it deletes anything, the app offers to save a copy of that device's meetings (recordings, transcripts and reports) to a place you choose, including your iCloud Drive. Each device saves its own. A subscription bought on our website is cancelled first. A subscription bought through the App Store is managed by Apple, so deleting your account does not cancel it: cancel it in your App Store settings. If you signed in with Apple, the app asks Apple to end that connection. Meetings you synced to your own iCloud are yours: they stay there unless you delete them, and we cannot reach them.

12. Recording other people

VoiceNear records the meetings you choose. You are responsible for telling the other participants and getting any consent the law requires where you and they are. In Online mode, their voices are also processed by our online service. The app shows that it is recording while a meeting runs.

13. Children

VoiceNear is not directed at children under 13, or under 16 in the EU and UK, and we do not knowingly collect their data.

14. Security

Information travels encrypted between your device and our service. We keep our service's logs free of meeting content and hash the network addresses we use against abuse.

15. Changes

We will post any change to this policy on this page, with a new date, before it takes effect.