SpeakStar — Privacy Policy
Last updated: 2026-08-05
SpeakStar is built by Sakura5 and is made for children, so this page is written for the parent. The short version: SpeakStar collects nothing, tracks nothing, shows no advertising, and makes no network requests of its own — there is no URLSession in the app, no backend, no analytics SDK, no advertising SDK, and no third-party SDK of any kind. It is free, has no in-app purchases, and needs no account. All 476 audio recordings are bundled inside the app, so it works with the internet switched off entirely — which is how we would suggest you let a child use it.
What the app handles
- Nothing about your child reaches us. The app's privacy manifest declares no collected data types and no tracking, and there is no server for it to send anything to.
- Your child's nickname, the English name they picked, their age, their level and their star count — stored locally in iOS UserDefaults on the device, and removed when the app is deleted.
- The most recent recording — written to a single temporary file on the device (“last-take.caf”) purely so your child can play back what they just said. It is overwritten by the next attempt and is never uploaded.
- Microphone audio while the mic button is held — passed to Apple's Speech framework on the device to be turned into text, then discarded (see “Speech recognition never leaves the device” below).
Speech recognition never leaves the device
We want to state this plainly rather than bury it, because a microphone in a children's app is the thing a parent should ask about first.
To score what your child says, the app has to convert their speech to text, and it uses Apple's own Speech framework on the device to do so. SpeakStar sets requiresOnDeviceRecognition = true unconditionally, on every device and on every attempt — there is no fallback path to server-side recognition.
On top of that, before the microphone is even switched on the app checks whether the device supports on-device recognition. If it does not, SpeakStar simply refuses to listen: the microphone is never opened, scoring is turned off, and the child is told the device cannot do it. We would rather the app score nothing than send a child's voice anywhere. There is therefore no code path, on any device, in which your child's voice reaches a server — ours, Apple's, or anyone else's.
If you would rather your child not use the microphone at all, the entire read-aloud section and all six songs work without it, and you can safely deny the Microphone and Speech Recognition permissions.
Because this is an app for children
SpeakStar does not knowingly or unknowingly collect personal information from children, because it collects no personal information from anyone. Specifically: there is no account, no sign-up and no email address; no advertising and no ad network SDK; no analytics and no third-party analytics SDK; no tracking, no advertising identifier and no cross-app or cross-site tracking of any kind; no profiling of your child and no behavioural targeting; no chat, comments, messaging, sharing or any other way for a child to communicate with a stranger; no user-generated content leaves the device; no links to social networks; and no data about your child is sold, rented, shared, disclosed or transferred to anyone. The nickname is a nickname — the app never asks for a real full name, a school, a birthday, a photo or a contact detail.
Permissions, and exactly when they are asked
Microphone and Speech Recognition are requested the first time your child taps the mic button — not at launch. The microphone records what they say so it can be compared with the target sentence and played back to them; speech recognition turns that audio into text so the words can be scored. The scoring itself is plain word-by-word text comparison written in Swift and run on the device.
Nothing else is ever requested: no camera, no photo library, no notifications, no location, no contacts, no health data.
Attribution
The 44 IPA reference recordings come from Wikimedia Commons under CC BY-SA 3.0 or in the public domain, recorded by phoneticians, and are credited inside the app (Bảng phiên âm → “Nguồn tư liệu”); they were converted to AAC without altering the content. IPA transcriptions were generated with espeak-ng (GPL). The sentence and word recordings and the six songs were generated for this app. The credits screen contains links to Wikimedia Commons and espeak-ng; tapping one opens your own browser, which is the only way the app can lead anywhere off the device.
What the scoring does and does not claim
SpeakStar reports which words of the target sentence were heard, which were heard as something close, and which were not heard at all, plus what the recogniser thought it heard, and awards 0 to 3 stars. It deliberately does not claim per-phoneme assessment, it is not an assessment of your child's ability, and it makes no educational or health claims.
Tóm tắt bằng tiếng Việt (dành cho ba mẹ)
SpeakStar không thu thập bất kỳ dữ liệu nào của bé. Ứng dụng không có tài khoản, không đăng nhập, không quảng cáo, không SDK thống kê, không SDK của bên thứ ba, và tự nó không gửi gì lên mạng. Toàn bộ 476 bản ghi âm nằm sẵn trong máy nên app chạy được khi tắt hẳn Internet. Tên gọi, tên tiếng Anh, tuổi, cấp độ và số sao của bé chỉ lưu trên máy; bản thu gần nhất chỉ ghi vào một tệp tạm để bé nghe lại và bị ghi đè ở lượt sau. App xin quyền Micro và Nhận dạng giọng nói đúng lúc bé bấm nút micro lần đầu. Việc nhận dạng luôn chạy ngay trên máy (requiresOnDeviceRecognition = true, không có đường lui): nếu máy không hỗ trợ nhận dạng trên thiết bị thì app không mở micro và tắt luôn phần chấm điểm, chứ nhất định không gửi tiếng của bé đi đâu cả. Nếu ba mẹ không muốn dùng micro, phần “Đọc to” và 6 bài hát vẫn dùng được bình thường. Không có chat, không chia sẻ, không theo dõi, và không bán hay chia sẻ dữ liệu của bé cho bất kỳ ai.
What we do NOT do
- We do not sell your data.
- We do not use your data for advertising or cross-app tracking.
- We do not offer accounts at all — no sign-up, no sign-in, no email address, no password, and no profile on any server.
- We do not collect any personal information from children, or from anyone else.
- We do not show advertising and the app contains no advertising SDK, no attribution SDK and no advertising identifier.
- We do not use analytics or crash reporting, and the app links no third-party SDK of any kind.
- We do not track your child across apps or websites, and we do not build a profile of them.
- We do not upload your child's voice, recordings, nickname, age or scores — the app has no networking code to upload them with.
- We do not provide chat, comments, messaging or sharing, so a child cannot be contacted by anyone through the app.
- We do not sell, rent, share or disclose any data to advertisers, data brokers, or anybody else.
Third parties
None. SpeakStar links no third-party SDK: no ad network, no analytics provider, no crash reporter, no subscription provider — the app is free with no in-app purchases, so even Apple's payment system is not involved. Apple's Speech framework is used only in its on-device mode, so nothing is sent to Apple either. The 44 IPA reference recordings are licensed from Wikimedia Commons (CC BY-SA 3.0 / public domain) and are bundled in the app; Wikimedia receives nothing about your use of it.
Children
SpeakStar does not knowingly collect personal information from children. SpeakStar is made for children roughly 4 to 10 years old and is offered in the App Store's Kids category. It collects no personal information from anyone, so it collects none from a child: no account, no advertising, no behavioural or contextual ad targeting, no analytics, no tracking or advertising identifier, no third-party SDK, and no data of any kind leaving the device. Parents and guardians with a question, or who want to check what the app stores, can write to support@sakura5.app; because nothing is ever transmitted to us, deleting the app removes everything it saved.
Contact
Email support@sakura5.app.