Privacy Policy
This Privacy Policy explains how MagicReader handles information. It applies to the MagicReader mobile apps (iOS and Android) and to the MagicReader web reader at leorivas.com/magicreader (together, “MagicReader”).
What we don’t do
- We don’t collect or store any personal data.
- No user accounts or sign-in.
- No analytics or usage tracking.
- No advertising or ad networks.
- No selling or sharing of your data.
- No third-party SDKs that collect data.
Camera
The camera is used only to recognize head, eye, and hand gestures for touch-free page turning. The video is processed on your device in real time to detect those gestures and is never recorded, saved, or transmitted. In the web reader, gesture detection runs entirely inside your browser (using Google’s MediaPipe running locally); the video never leaves your device. The camera is optional — MagicReader works without it, and you can turn it off at any time. See Face data below for exactly what is computed from the camera.
Face data
The iOS app’s optional “Wink” control uses Apple’s ARKit face tracking (the TrueDepth API) to detect whether each of your eyes is open or closed, so you can turn pages without touching the screen.
- What is processed: numeric eye-openness values (ARKit blendshape coefficients) and head orientation, computed frame by frame. The app never creates, stores, or transmits face images, depth maps, or any biometric template (“faceprint”). No face recognition or identification is performed.
- Where it is processed and stored: entirely on your device, in memory only. Face data is used for the instant of the current camera frame to decide whether a wink or head movement occurred, and is discarded immediately. It is never written to disk.
- Retention: none. No face data persists after the frame is processed — closing the feature, the book, or the app leaves nothing behind.
- Sharing: none. Face data is never shared with the developer or any third party, and never leaves your device. MagicReader contains no analytics, advertising, or tracking SDKs.
- Deletion: because nothing is retained, there is nothing to delete; disabling Wink control or camera access (Settings → MagicReader → Camera) stops all face processing immediately.
- Web reader: the same rules apply. The browser version computes equivalent numeric values (eye-openness scores, head angle, and hand landmark positions) locally with Google’s MediaPipe — in memory, frame by frame, never stored, never transmitted. Turning the camera off (the ✕ on the camera panel, or your browser’s camera permission) stops all processing immediately.
Microphone & speech recognition
The microphone is used only to recognize a small set of reading commands (for example, “next page”). MagicReader never records, saves, or transmits your audio.
- In the mobile apps, speech recognition is performed by your device’s operating system. MagicReader requests on-device recognition where your device supports it. Any speech processing performed by the operating system itself is handled by your device maker (Apple or Google) under their own privacy policies.
- In the web reader, voice commands use your browser’s built-in Web Speech feature. Some browsers (for example, Google Chrome) may process this audio on the browser provider’s servers rather than on your device; that is governed by your browser’s privacy policy. Voice control is optional and off by default.
The microphone is optional — MagicReader works without it.
Network
MagicReader connects to the internet only to download content you request:
- Free public-domain books from sources such as Project Gutenberg (gutenberg.org) and Aozora Bunko (aozora.gr.jp).
- Read-aloud voices: the standard voice is your device’s built-in speech synthesis. The optional Natural voice is a speech model (Kokoro) downloaded once — in the mobile apps from our model host, in the web reader from a public model CDN (Hugging Face / jsDelivr) — and it generates speech entirely on your device. The text you read and the audio it produces are never uploaded.
- Web reader only: the gesture-recognition model files are downloaded once from a public code CDN (jsDelivr / Google) so detection can run locally in your browser.
These requests contain no personal information and no account identifiers. As with any internet request, the third-party servers you download from may automatically receive your device’s IP address and standard technical details (such as browser or OS type) as a normal part of how the internet works. That information is handled by those third parties under their own privacy policies. MagicReader does not add any identifying information to these requests.
Your content and settings
Your library, reading positions, and settings are stored on your device (and in your own device backups, such as iCloud or Android/Google backups, if you use them). We have no access to them.
Children’s privacy
MagicReader does not collect personal information from anyone, including children. It contains no ads, no data collection, and no external links to user-generated content, and is suitable for general audiences.
Permissions summary
- Camera — optional; used only for on-device gesture recognition.
- Microphone / Speech recognition — optional; used only for voice commands.
You can grant or revoke these at any time in your device or browser settings, and MagicReader remains usable without them.
Changes to this policy
If this policy changes, we’ll update this page and the “Last updated” date above.
Contact
Questions about privacy? Email fusioncalc@gmail.com.