MimiSeka is an iOS app that uses AI to interpret scenes from photos and videos, generating literary narratives and summary text. The generated text can be embedded as photo captions (ALT text) or read aloud.
- MimiSeka Scia: Generates scene summaries and narratives from your photos. You can also talk with the AI about a photo and have a diary entry written from that conversation. The generated text can be saved as ALT text on the photo or read aloud. What you write collects in an on-device shelf, and can optionally be published as a link or QR code.
This policy describes how data is handled in this app. It applies to users in the European Economic Area (EEA), the United Kingdom, and Switzerland.
1. Data Controller
The data controller responsible for your personal data is:
MimiSeka Project
Email: privacy@mimiseka.jp
2. Data We Collect, Transmit, and Legal Basis
Under the General Data Protection Regulation (GDPR), we are required to inform you of the legal basis for processing your personal data. The table below summarises the data we collect, why we collect it, and the legal basis for each.
| Data | Purpose | Legal Basis |
|---|---|---|
| Photos (sent to generative AI services) | Generate summaries and narratives | Consent (you choose to use the feature) |
| Initial survey (vision status, how you see, age, read-aloud preference) with device model, iOS version, language and region | Optimise app settings; statistics for service improvement (no identifier attached) | Legitimate interest (service improvement) |
| Device identifier (identifierForVendor) | Usage limits (weekly in Scia); recording purchased allowances; confirming genuine app requests and preventing abuse | Legitimate interest (service stability) |
| Location data (optional) | Photo metadata; context for stories and conversations; on-device map and register | Consent (iOS system permission prompt) |
| Usage data, with survey attributes, device model, coarse location (country and region), and a daily-rotating pseudonymous code | App improvement and analytics | Legitimate interest (service improvement) |
| Text of a story you place at a location (sent to OpenAI's Moderation API) | Check content before publication | Legitimate interest (safety of recipients); legal obligation (preventing distribution of illegal content) |
| Pseudonymised author / reporter identifier (salted hash), block records | Let recipients block an author; identify repeat offenders | Legitimate interest (safety of recipients and moderation) |
| Report reason and note | Review reported content and decide whether to restore or delete it | Legitimate interest (safety of recipients and moderation) |
| X (Twitter) authentication token | Optional posting to X | Consent (you choose to connect) |
| Conversation, diary, and note text (sent to generative AI services) | Write diary entries and stories from what you say | Consent (you choose to use the feature) |
| Conversation memory and writing-style notes (kept on device; sent to generative AI services as context) | Let conversations and diary entries reflect how you speak and write | Consent (you choose to use the conversation and diary features; you can delete the memory at any time) |
| Text to be read aloud (sent to speech synthesis services) | Narrate stories and diary entries | Contract performance (providing the read-aloud feature you use) |
| Grid-cell number of the capture point (about 500 m; no coordinates or identifier) | Show how many photos were taken around here | Legitimate interest (a feature of the capture screen) |
| Published content (text, place name, narration audio, and the photo — for QR codes only if you opt in; always for "Send as link") | Make a story listenable to anyone holding the link ("listenable photo") | Explicit consent (you choose to publish, and may revoke at any time) |
| Story left at a place (text, audio, capture date and time, the place) | Show it anonymously to other users who visit that place | Explicit consent (you agree before the first time, and may withdraw at any time) |
| Purchase receipt (signed by Apple) and device identifier | Verify in-app purchases and record the allowance bought | Contract performance (necessary to deliver what you paid for) |
Photo Data
To generate summaries and narratives for your photo captions, photos you take or select are sent to external generative AI services (OpenAI and Anthropic). How data is handled after transmission is governed by each service's privacy policy.
- MimiSeka Scia: Generated summaries and narratives are embedded as metadata in the photo and saved to your device's photo library (in the "MimiSeka" album). This feature can be disabled in Settings.
Conversation, Diary, and Note Text (Scia)
The text you type or speak while talking about a photo, any note you attach to a photo, and the material used to write a diary entry are sent to third-party generative AI services in order to produce the writing. Once transmitted, that data is handled according to each service's own privacy policy.
To make conversations and diary entries feel natural, the app also sends, as context each time it generates a conversation reply or a diary entry, the "conversation memory" kept on your device (notes distilled from past conversations about how you speak and what you tend to care about) and notes on your writing style (its characteristics and short samples). Building the conversation memory also sends what you said to the generative AI service.
The conversation record itself, and these notes, are kept on your device (see 4. Data Stored on Your Device).
Initial Survey
On first launch, the app asks for the following information:
- Vision status
- How you see (e.g. blurred, glare, narrow field of view; optional)
- Age
- Auto-read-aloud setting (optional)
This information is used to optimise app settings. To improve the service, it is also stored on our servers together with the settings chosen automatically from your answers and your device model, iOS version, language, region, and whether VoiceOver is on. This record contains no device identifier or anything else that identifies you or your device, and it is never used to identify any individual.
Device Identifier
A device-specific identifier assigned by Apple (identifierForVendor) is sent to our server with each request. It is used only to:
- Enforce usage limits (weekly in Scia)
- Record usage allowances you have purchased
- Confirm that requests come from the genuine app and prevent abuse
This identifier changes when the app is reinstalled and is not used for cross-app tracking or advertising purposes.
Location Data (Optional)
Location data is only collected when you grant the app location access through your device's iOS system settings. When enabled, location data is used for:
- Adding location metadata to photos (when saving to album in Scia)
- Providing contextual information for stories and conversations
- Showing where a photo was taken on a map inside the app
- Building an on-device register of the regions you have visited
- Retrieving place and venue names for the capture point via Apple Maps
- Retrieving the weather at the capture point and time via Apple WeatherKit
Precise location data (latitude and longitude) is sent to our servers only if you use the "Leave a Story at This Place" feature described below. Otherwise, location-related information is sent only in the following forms:
- To generative AI services — the municipality-level place name, venue name, and the weather at capture time may be sent as text, as context for stories and conversations
- To our servers — to show "photos taken around here" on the capture screen, the number of the roughly 500 m × 500 m grid cell containing the capture point is sent. Latitude and longitude are not sent; only a count of photos per cell per day is recorded, with no device identifier or anything else that identifies your device
Place names, venue names, and weather are retrieved through Apple's services and do not pass through our servers.
Content You Publish as a "Listenable Photo" (Optional)
MimiSeka Scia lets you turn a story or diary entry into a link or QR code so that people without the app can listen to it. Only when you use this feature, the following is stored on our servers (AWS Tokyo region) and becomes accessible to anyone who has the link:
- Text — the story or diary entry, its title, capture date, the place name, and other display information
- Narration audio — the synthesized audio file
- Photo — when publishing as a QR code, stored only if you turn on "Photo" in the publishing settings (off by default). When you use "Send as link", the photo is always stored so the recipient can listen alongside it. In both cases the JPEG with the story embedded is stored as-is
Publication is time-limited; expired content becomes unreadable and is deleted automatically.
- QR code — you choose the period when you publish. Before expiry, you can revoke publication at any time via "Stop QR publishing" (or "Stop all publications" in Settings); once revoked, the link no longer opens
- Send as link — the period is fixed at 7 days. Because the app keeps no record of links you send, a sent link cannot be revoked before it expires. Please check the recipient and content before sending
The legal basis is your explicit consent, given at the moment you publish. Withdrawal does not affect the lawfulness of processing before withdrawal.
Our servers count how many times published content has been played or viewed and report that count back to you. We do not record who viewed it.
"Leave a Story at This Place" (Optional)
MimiSeka Scia lets you leave a story at the place where the photo was taken. If you use this feature, your story (text, narration audio, capture date and time, and the place where it was left) is stored on our servers and shown anonymously to other users who visit that place. The publication period is one year.
The legal basis is your explicit consent, which you give before the first time you use the feature. Your name and display name are never shown. You may withdraw at any time from within the app; withdrawal stops new deliveries, though the story may remain on devices that already received it.
Before publication, the text is sent once to an automated content check (OpenAI's Moderation API). Text judged inappropriate is neither stored nor published (so that we can review the decision, the result and up to the first 200 characters of the text are sent to the operator). Photos are never stored by this feature, and recipients see only the month and year of capture. The legal basis for this check is our legitimate interest in keeping the service safe for the people who receive these stories, and our obligation as a platform to prevent the distribution of illegal content.
Stories can be reported by other users. A reported story stops being delivered immediately; we then review it and decide whether to restore or delete it. The reason selected by the reporter, and any note they write, are stored for up to 400 days for audit purposes.
For this feature only, we store a pseudonymised identifier for the author and for the reporter, derived from the publishing identifier held on your device by a salted, irreversible hash. It is used solely to exclude an author's stories from delivery when a recipient blocks them, and to identify repeat offenders. It is never displayed and never passed to other users. Authors and reporters are hashed with separate domains, so the two cannot be matched against each other. Because a pseudonymised identifier is still personal data under the GDPR, your rights of access, erasure, and objection apply to it; contact us and we will act on the request.
Push Notifications (Optional)
"Letters from MimiSeka" are notifications created on your device; they are not sent from our servers. Notices delivered through iCloud (such as reactions to something you shared) use Apple's Push Notification service (APNs). The legal basis is your consent, given through the iOS permission prompt, and you can revoke it at any time in iOS Settings.
iCloud (CloudKit) Storage and Sync
The records collected on your shelf are stored and synced in your own iCloud. This includes story text and titles, capture notes, emotion tags, narration audio, the location where each photo was taken, the weather at capture time, the AI's analysis of the photo, and your chosen writing style and vision settings. The photo images themselves are not sent — they stay in your device's photo library, and only an identifier for each photo is stored in iCloud. This happens entirely within Apple's services and does not pass through our servers. Data held in iCloud is governed by Apple's privacy policy, with Apple acting as the controller for that storage.
In-App Purchases
If you buy a subscription or a ticket pack, we verify the signed purchase receipt issued by Apple on our servers and record the resulting allowance against your device identifier. We never receive your payment details, such as credit card numbers.
Usage Data
To improve the app, the following usage data is recorded and sent to our servers:
- Number and completion rate of story generations, the writing style chosen, and generation time
- Feature usage frequency, and whether narration was listened to the end
- Error occurrences
- Accessibility feature usage (VoiceOver, etc.)
For analysis, each record carries attributes: your initial survey answers (vision status and age), language and region, device model, and a coarse location (country and state/prefecture).
Records do not contain the device identifier itself. However, to link records from the same day, each carries a code that changes every day (an irreversible value derived from the device identifier and the date). Because the code changes with the date, the same device cannot be followed across days. This is pseudonymous rather than anonymous data. It is used solely for service improvement and never to identify any individual.
Because this is pseudonymised data, it remains personal data under the GDPR. The legal basis is our legitimate interest in improving the service; you have the right to object.
Data Retention
Content published as a "listenable photo" is retained until the period you chose when publishing (30 days by default) for QR codes, or for 7 days when sent via "Send as link", after which it is deleted automatically. A story left at a place is retained for one year. Report records are retained for up to 400 days, and block records until they are removed. Withdrawing a QR publication or a story left at a place in the app makes it unreadable from then on. Per-cell photo counts are retained for up to 5 years. Usage data and device identifiers are retained for the duration of service operation and deleted when the service is discontinued. Initial survey data is retained as statistical information for the duration of service operation.
3. Third-Party Services (Data Processors)
We use the following third-party services to provide app functionality. These services act as data processors on our behalf or as independent controllers for data transmitted to them.
Generative AI Services
The following generative AI services are used for photo analysis and narrative generation. All communications are routed through our server over encrypted connections (HTTPS).
- OpenAI — Photo analysis, text generation, content check for "Leave a Story at This Place" (Moderation API)
- Anthropic (Claude) — Photo analysis, text generation
How data is handled after transmission is governed by each service's privacy policy:
Text-to-Speech Services
The following services may be used to convert generated text to audio. Only the text to be read aloud is sent (including text you wrote or edited yourself); no photos or user-identifying information is included.
- OpenAI — Speech synthesis for summaries and narratives
- Aivis Cloud — Speech synthesis for summaries and narratives (when using the natural Japanese voice)
X (formerly Twitter)
Posting to X is optional. If you choose to use this feature, you connect your account via X's secure account authorisation. The app requests only the permissions needed to post (X requires permission to read posts and your account information in order to post). The app never reads your timeline or followers. The photo and text you post are sent directly from your device to X and do not pass through our servers. Authentication credentials are securely stored in the device's secure credential storage (Keychain) and deleted upon logout.
4. Data Stored on Your Device
Scia
Generated summary and narrative text can be embedded as a description field stored inside the photo file (IPTC caption) and saved to the "MimiSeka" album in your device's photo library. This feature can be toggled on or off in Settings. Saved photos may include the following embedded data:
- Summary/narrative text (as ALT text) and title
- Emotion tags (if selected), weather and time of day at capture, and a composition rating
- Place name and location data (if location services are enabled)
- A marker showing the photo was made with MimiSeka, and a per-photo identifier
If you share the photo from the Photos app, this information travels with it.
We also store, on your device, the record of your conversations about photos, the "conversation memory" distilled from them (notes on how you speak and what you tend to care about), and notes on your writing style. These are not synced to iCloud (CloudKit). The conversation memory and writing-style notes are also excluded from device backups (the conversation record is included in device backups). They are, however, sent to generative AI services as context when conversations and diary entries are generated (see Conversation, Diary, and Note Text). You can review and delete the conversation memory from "Review conversation memory" and "Delete conversation memory" in the "…" menu on the conversation screen, and "Erase all data" in Settings deletes all of it.
Voice Recordings Attached to Photos
In Scia you can record your own voice and attach it to a photo (before or after taking it, or later from the chapter screen). Recording your voice is the only feature that uses the microphone.
Recorded audio files are stored in the app's own storage on your device. They are never transmitted to our servers, and they are never transcribed or analyzed. Audio is not included when you publish a story as a QR code or link, when you share it with someone, or when you leave a story at a place.
Because these recordings are part of your device backup (such as an iCloud backup), they carry over when you move to a new device. You can remove the audio from an individual chapter inside the app, and "Erase all data" in Settings deletes all of them.
5. Data Security
- All external communications are encrypted via HTTPS
- Service credentials are managed on our server, not stored in the app
- Apple's App Attest framework is used to verify that requests come from genuine, unmodified copies of the app
- X credentials and the App Attest key identifier are stored in the device's secure credential storage (Keychain)
6. Sharing with Third Parties
Except for the data processors and services described above, we do not sell, rent, or otherwise provide your data to third parties.
7. International Data Transfers
Your data may be processed by third-party generative AI and speech synthesis services in countries outside the EEA/UK that may not offer the same level of data protection. All transfers are conducted over encrypted connections (HTTPS) via our server. Where personal data is transferred outside the EEA/UK, we rely on the European Commission's adequacy decisions or standard contractual clauses as appropriate to ensure adequate protection of your data.
8. Children's Privacy
This app is not directed at children under the age of 16 in the EEA/UK, but it may be used by children with visual impairments. Survey responses and usage data are collected for all users regardless of age without being linked to names, contact details, or other personal information (see "Usage Data" for how usage data is handled). Device identifiers are sent to our server only for the purposes described under "Device Identifier" and are not linked to any personal information. Core app features (photo analysis, narrative generation, and read-aloud) remain fully available regardless of age.
That said, publishing a "listenable photo" and leaving a story at a place are features where what you write can be seen by other people. If a child uses the app, a parent or guardian should review these features first. Any publication can be withdrawn.
If we become aware that we have inadvertently collected personal data from a child without appropriate consent, we will take steps to delete such data promptly. If you believe a child has provided us with personal data, please contact us at privacy@mimiseka.jp.
9. Your Rights Under the GDPR
If you are in the EEA, UK, or Switzerland, you have the following rights regarding your personal data:
- Right of Access — Request a copy of the personal data we hold about you
- Right to Rectification — Request correction of inaccurate or incomplete personal data
- Right to Erasure ("Right to be Forgotten") — Request deletion of your personal data
- Right to Restriction of Processing — Request that we limit how we use your data
- Right to Data Portability — Request your data in a structured, commonly used, machine-readable format
- Right to Object — Object to processing based on legitimate interest
- Right to Withdraw Consent — Where processing is based on consent, you may withdraw it at any time without affecting the lawfulness of prior processing
To exercise any of these rights, please contact us at privacy@mimiseka.jp. We will respond to your request within 30 days.
Right to Lodge a Complaint
You have the right to lodge a complaint with a supervisory authority in the EU/EEA member state of your habitual residence, place of work, or place of the alleged infringement. For the UK, you may contact the Information Commissioner's Office (ICO).
Additional Controls
- Disable location services, album saving, and other features at any time from the in-app Settings
- Disconnect X integration by logging out, which deletes your authentication credentials
- Revoke a "listenable photo" published as a QR code at any time via "Stop QR publishing" in the app or "Stop all publications" in Settings; once revoked, the link no longer opens (content sent via "Send as link" cannot be revoked and is deleted automatically after 7 days)
- Withdraw a story you left at a place at any time via "Stop placing" in the app
- Review and delete the conversation memory at any time from the "…" menu on the conversation screen
- Delete all on-device data at any time via "Erase all data" in Settings
- Delete all on-device data by uninstalling the app (excluding photos already saved to your photo library). Withdraw anything you have published as a "listenable photo" or left at a place before uninstalling, or contact us
10. Changes to This Policy
If we make changes to this policy, we will notify you through the app or our website. For significant changes, a notification will be displayed when the app launches.
11. Contact Us
If you have any questions or requests regarding privacy or wish to exercise your rights, please contact us:
Data Controller: MimiSeka Project
Privacy enquiries: privacy@mimiseka.jp
Reports, problems, and general enquiries: support@mimiseka.jp (also reachable from Settings → About → Contact us in the app)