Effective October 27, 2026 Applies to the WristEcho (腕上回声) iPhone app, its Apple Watch app, its watch face complication and this website
In short: WristEcho has no servers and collects no data. Recording, transcription, voiceprints, search and AI summaries all happen on your own devices. The developer cannot see or access any of your content.
Who is responsible
WristEcho is developed and published by an independent developer, the developer shown on the App Store. "We" means the developer and "you" means the person using the app. For any privacy question, email wty2003728@gmail.com.
We collect no data
There is no account. You never sign up or log in.
We run no servers. Your recordings, transcripts, voiceprints and reports are never sent to us.
The app contains no third-party analytics, advertising, crash-reporting or tracking code, and it does not track you across apps or websites.
We do not sell, rent or share personal information, because we don't have any.
The app's App Store privacy label is "Data Not Collected".
Data processed only on your devices
Data
Where it comes from
Where it is kept
Recordings
Audio recorded after you tap Start on Apple Watch or iPhone, or audio files you import
iPhone. The watch only holds segments until they reach the iPhone
Transcripts
Your iPhone turns the audio into text on the device
iPhone
Speaker labels, voiceprints and voice features
See the next section
iPhone
Search index
To search by meaning, your iPhone turns text into numerical vectors
iPhone, excluded from backups and rebuilt from the transcripts when needed
Daily and weekly reports, titles, answers
Generated on the device by Apple Intelligence, a local model you downloaded, or extract mode
iPhone
To-dos
Taken from your reports; you can edit, check off and delete them
iPhone. If you tap Add to Reminders, they are written to your default Reminders list and synced by the account that list belongs to (iCloud, or an Exchange, Google or other account you added)
Settings
Your choices in the app, such as transcription language and cleanup
iPhone and Apple Watch
If you use iCloud Backup or back up your iPhone to a computer, the transcripts, voiceprints, reports and other data above are also included in your own backup (audio is left out by default). See "iCloud and computer backups" below.
Transcription uses Apple's built-in offline speech recognition or a speech model bundled with the app. Speaker separation and voiceprints use models bundled with the app, and search by meaning uses Apple's NaturalLanguage framework. All of these run on the device.
Voiceprints and other people's voices
Recordings capture the people around you. When you talk with others, let them know you are recording and follow the recording laws where you are; some places require everyone present to agree. Whether and how you inform people and obtain consent is your responsibility. The app asks you to confirm this before first use, and a clear indicator is shown on the watch and iPhone while recording.
What a voiceprint is. A voiceprint is a set of numbers calculated from a sample of someone's speech, used to tell whether the same person is speaking in later recordings. It is not the recording itself.
Voice features. So that people can be recognized later, each transcript stores a voice feature (also a set of numbers) for every speaker in it. For a speaker who isn't recognized, the feature is deleted together with that segment's audio; segments with only strangers or background sound are cleaned up after 14 days by default. Removing a person deletes the voice features tied to them. When you confirm that someone knows and agrees to be remembered, earlier recordings that still hold their voice are labeled with their name too.
How voiceprints are created.
You read a short text to your iPhone for about 20 seconds.
A family member or colleague reads the same text next to you. Before recording, the app reminds you to get their consent first.
When you label a speaker in a recording as a particular person, only that segment is labeled by default and no voiceprint is extracted. Only after you confirm that the person knows and agrees to be recognized does the app extract that voice's features from the recording and add them to the person's voiceprint so they can be recognized later. Labeling a speaker as yourself refines your own voiceprint directly.
How voiceprints are used. Only to label who is speaking in your recordings. Voiceprints and voice features stay on your iPhone and never leave your devices except inside your own iCloud or computer backups. They are never sent to us or anyone else, and are never used to infer emotion, health, age, gender or any other personal characteristic.
How to delete them. In Settings › People & Voiceprints you can delete a person or re-record a voiceprint. Deleting a person keeps the recordings and text but stops labeling that person's name, and deletes the voice features tied to them. Delete All Data removes every voiceprint and voice feature.
In some places voiceprints count as biometric or sensitive personal information, for example under the EU General Data Protection Regulation, China's Personal Information Protection Law and Illinois' Biometric Information Privacy Act. Get a person's explicit consent before recording their voiceprint.
Where data is stored and how it is protected
Data is kept in the app's private storage on your iPhone. Other apps cannot read it, and it does not appear in the Files app.
Data is covered by iOS Data Protection and becomes available after the first unlock following a restart, so recording and transcription can continue while the phone is locked.
On Apple Watch, audio is held in 10-minute segments and sent to your paired iPhone through Apple's Watch Connectivity, never through us. Each segment is deleted from the watch once the iPhone confirms it arrived.
You can turn on App Lock in Settings › Privacy. The app then asks for Face ID, Touch ID or your passcode every time you open it or return after more than a minute, and hides its content in the app switcher.
iCloud and computer backups
WristEcho does not use iCloud sync or store data in any cloud. If you use iCloud Backup or back up your iPhone to a computer, iOS backs up app data under Apple's rules:
Audio is excluded from backups by default. Settings › Storage › Keep Audio Out of iCloud Backup is on by default; turn it off to include audio. Apple describes this setting as a hint to the system rather than a guarantee.
Transcripts (including their voice features), reports, to-dos and voiceprints are included in device backups. They come back when you restore a new iPhone from that backup.
The search index is not backed up. After a restore, the app rebuilds it from the transcripts.
Downloaded local models are never backed up.
Apple stores backups encrypted under your Apple Account, end-to-end encrypted if Advanced Data Protection is on. We have no access to your backups.
When the app goes online
WristEcho has no servers of its own. It goes online only in these cases, and none of them uploads your recordings, transcripts or voiceprints:
Speech recognition assets. The first time you transcribe a language, iOS downloads the offline recognition files for it from Apple.
Language assets. When the search index is first built, iOS downloads the language files the NaturalLanguage framework needs from Apple.
The Apple Intelligence model is downloaded and managed by iOS, not by the app.
Local model download, only after you confirm it. The app downloads public model files from huggingface.co, falling back to the mirror hf-mirror.com. The requests contain only the file addresses, with no account, cookies or content of yours. As with any website, those servers see your IP address and basic information the system adds automatically, such as the OS version. Both sites are run by third parties with their own privacy policies.
App Store region. The app reads your App Store country or region through StoreKit to decide whether generative AI is offered.
Web pages and email. The privacy policy, terms and help pages open in Safari. Contact Us opens your own Mail app, and you decide whether and what to send.
How AI is used
Apple Intelligence: on supported iPhones, reports, titles and answers are generated on the device with Apple's Foundation Models framework. Private Cloud Compute is not used. Apple Intelligence itself is covered by Apple's privacy policy.
Local model: an open-source model you choose to download (Qwen3, Apache 2.0 license) runs on your iPhone with MLX.
Extract mode: no generative model at all; it picks original sentences from your transcripts.
WristEcho uses no third-party AI service and needs no API key. AI-generated content can be wrong, so check important details against the original audio. If WristEcho is later offered on the China mainland App Store, copies downloaded there will not offer generative AI features and will use extract mode only.
Purchases
WristEcho is a paid download from the App Store, with no in-app purchases. Apple processes payment; we never receive your name, Apple Account or payment details. The sales reports Apple gives developers are aggregated and do not identify you.
Crash reports and diagnostics
The app contains no crash-reporting tool. Only if you turn on Share With App Developers in iPhone Settings › Privacy & Security › Analytics & Improvements does Apple provide us with crash logs and aggregated usage data. Apple collects that data, and it does not contain your recordings or transcripts.
If something goes wrong, Settings › Help & Support › Diagnostics creates a diagnostics file: app and OS version, device model, settings, counts such as recording segments and processing queues, and the crash and performance reports iOS gives the app. It contains no recordings, text, names or voiceprints. You see everything in it first, and it leaves your iPhone only if you share it yourself.
Retention and deletion
Your data stays on your devices until you delete it.
Automatic audio cleanup is on by default and adjustable in Settings › Storage: audio of segments with only strangers or background sound is kept 14 days by default (1 to 90 days); audio with people you know is kept 90 days by default (or 30 days, 1 year or forever). Until you record your own voiceprint, unrecognized speech follows the people-you-know period. Starred segments keep their audio, audio that hasn't been transcribed yet is never cleaned up automatically, and audio of segments with no speech is deleted right away. Cleanup removes audio only; text and search stay.
When storage runs low: with less than 2 GB free, the app first removes audio with only strangers or background sound, then other audio older than 30 days. Starred segments and all text are never touched.
Manual deletion: you can delete a whole recording, a single 10-minute segment, a person's voiceprint, a downloaded local model and the sample data.
Delete All Data in Settings removes recordings, transcripts, voiceprints, reports, to-dos and search history from this iPhone and keeps your settings. Segments still on the watch that have not reached the iPhone will arrive later.
Deleting the app removes all of its data from the device. Existing iCloud or computer backups follow Apple's rules, and you can manage or delete them in system settings.
To-dos you added to Reminders and files you exported or shared are under your control; delete them where they are.
Export
Settings › Storage › Export All Text collects all transcripts, by day, into one text file you can save to Files or send to yourself. You can also share the text of a single recording or report, or the audio and text of a single segment. Exported files are under your control.
Children
WristEcho is not designed for children under 13, or under the age set by local law. We collect no personal information from anyone, including children. If a parent is concerned about a child's use, all data is on the device and can be deleted as described above.
On Apple Accounts where the law requires it (for example in Texas), the app asks the system once for your age range (Apple's Declared Age Range). It does not store your age or send it to us.
Your rights
Because your data lives only on your devices and in your own backups, you can view, export, correct (for example, edit a line of a transcript, a report's title and summary, or a speaker label) and delete it directly in the app. We hold no data about you, so we cannot look it up, correct it or delete it for you, and we cannot recover lost data.
If you email us, we receive your email address and what you write. We use it only to reply, never for marketing, and we don't pass it to anyone. Our support mailbox is hosted by Gmail, so messages pass through Google's servers. You can ask us to delete our correspondence at any time.
This website
This website is a set of static pages hosted on GitHub Pages. It uses no cookies, no analytics and no external fonts or scripts. GitHub may log visitors' IP addresses for security and operations; see GitHub's Privacy Statement.
Changes and contact
If this policy changes, we will update this page and the effective date above, and point out significant changes in the app's release notes. Before adding any feature that would collect data, we will update this policy and the App Store privacy label first.
