Privacy Policy · SentenceDeck
Effective Date
October 5, 2026
Scope
This policy applies to the iOS app SentenceDeck (the “App”) and the services used to provide its optional premium voice and scene-dialogue features.
Quick Summary
- No SentenceDeck account or sign-up is required.
- System voice, microphone recordings, practice progress, articles, saved dialogues, and generated audio are handled on your device.
- Premium voice sends the text you choose and its voice settings through a Cloudflare Worker to Google’s Gemini speech model on Google Cloud (Vertex AI), a generative AI service, after you agree in the App.
- Scene dialogues are written by AI on our server. Preset scenes send only the scene, languages and level you choose. If you describe your own scene, that description is sent through our server to Google’s Gemini service (Google Cloud Vertex AI) to write the dialogue — only after you agree in the App.
- You can report any AI-written dialogue. A report contains the dialogue’s ID and the reason you choose, not the dialogue text.
- Apple processes In-App Purchases. SentenceDeck uses limited transaction and character-wallet records to validate purchases and prevent duplicate charging.
- We do not use advertising SDKs, sell personal data, or track you across apps and websites.
Data We Process
1) Content you choose for premium voice
When you request premium audio, the App processes:
- The sentence text you submit for speech generation.
- Language, voice, speaking style, and generation settings required to create the audio.
- Request identifiers needed to authorize the request, reserve characters, and prevent a completed request from being charged twice.
The App sends this information to a Cloudflare Worker operated for SentenceDeck. The Worker sends the sentence text, with its voice and speaking-style settings, to Google’s Gemini speech model on Google Cloud (Vertex AI), a generative AI service, which returns the audio. In SentenceDeck 2.0 a sentence from a scene dialogue may also carry a short English description of how it should be spoken (for example “flustered and apologetic”), written by the dialogue generator. The generated audio is returned to the App and saved locally. The App asks for your agreement before the first premium request; SentenceDeck 2.0 asks again even if you agreed in an earlier version, because the service changed.
System voice uses Apple’s on-device speech engine and does not send sentence text to the SentenceDeck premium-voice service.
2) Scene dialogues (SentenceDeck 2.0)
When you start a preset scene, the App sends the scene ID, practice language, interface language, level, the Standard or Detailed option and a request identifier to a Cloudflare Worker operated for SentenceDeck. The Worker asks Google’s Gemini models on Google Cloud Vertex AI to write the dialogue and returns it to the App, where it is saved on your device. Preset dialogues may be cached on the service for up to 7 days and shared between users, because they contain no personal data. The built-in offline scenes are included with the App and are not requested from the service.
When you describe your own scene, the description you type (up to 200 characters) is sent the same way, and only after you have agreed in the App. We ask you not to include names, contact details or other personal information, and descriptions containing web addresses, email addresses or phone numbers are refused before they reach the AI service. We do not store your description. The generated dialogue is kept on the service for up to 24 hours under a one-way hash of your wallet and the description, only so that a retried request is not charged twice. You can withdraw your agreement in Settings › Generation.
To enforce the daily allowance of free Standard dialogues, the service counts how many dialogues each wallet started per day; the counts are deleted automatically about 8 days later. Detailed dialogues reserve characters while the dialogue is written and charge them only when it is delivered; the charge appears in your wallet history. If a Detailed dialogue is cancelled while it is being written, the reserved characters are returned; the service counts these cancellations per wallet per day to limit abuse, and those counts are also deleted automatically about 8 days later.
Service logs for dialogue generation record the option, model, scene ID, languages, level, outcome, size and timing. They do not contain your description or the generated dialogue.
If you report a dialogue, the service stores the dialogue’s ID, the reason you selected, your wallet identifier and the time. Because the dialogue text is not stored, we use reports to see how often each scene and dialogue is flagged and to adjust the prompts and safety settings that write them; a report cannot be used to read back the conversation.
3) Purchases, wallet records, and service diagnostics
Apple processes payments. SentenceDeck receives transaction information needed to validate a purchase and credit the correct character pack. We do not receive your full card or bank details.
To secure premium requests and keep the character balance accurate, the service processes limited records such as:
- App transaction and request/session identifiers.
- Character reservation, commit, release, and wallet-balance records, including charges for Detailed dialogues.
- Product identifier, transaction status, and credited character amount.
- Technical diagnostics such as endpoint, voice/model selection, status, timestamps, and error context.
Infrastructure providers may process standard network information, such as an IP address and user agent, as part of security and operational logging.
4) Microphone and practice data
If you grant microphone access, SentenceDeck records your voice so you can compare it with the reference sentence. The recording and practice state are processed locally. SentenceDeck does not send your microphone recording to its services and does not use it for AI pronunciation scoring.
5) Local articles, dialogues, and audio
Your imported or pasted articles, saved scene dialogues, practice progress, reference audio, and microphone recordings are stored in the App’s local container as needed for the feature. SentenceDeck does not provide developer-operated account sync for this library.
How Premium Voice and Scene Data Flow
- App → Cloudflare Worker for request authorization and billing control.
- Cloudflare Worker → Google Cloud Vertex AI (Gemini speech model) for speech generation, or Gemini models for writing scene dialogues.
- Generated audio or dialogue → App, where it is saved on your device.
Authentication and character-wallet records, daily dialogue counts, and dialogue reports may be stored in service infrastructure, including Google Cloud Firestore, to validate purchases, enforce the free daily allowance, and maintain the wallet across supported devices associated with your Apple transaction identity.
Device Backup, Export, and Import
Premium audio is eligible for iCloud or computer device backup when that backup is enabled and has sufficient space. Free system-voice audio can be regenerated and may be excluded from device backup.
Device backup is controlled by Apple and the device owner. Premium audio returns only as part of a compatible full-device restore. Deleting and reinstalling SentenceDeck alone does not restore that audio.
When you manually export audio or create a backup:
- SentenceDeck creates a temporary ZIP or merged audio file and presents Apple’s system share or document controls.
- You choose the destination, such as Files, iCloud Drive, AirDrop, email, or another compatible service.
- SentenceDeck does not receive access to the destination account and does not upload the file to a developer-operated storage service.
- Temporary export files are removed from the App’s scratch space when the export flow ends.
When you import a SentenceDeck ZIP, the package is copied to temporary App storage, checked for compatible content and audio, and processed locally. The temporary copy is removed when the import session ends. Compatible content and audio can be restored; practice progress is not included. Audio generated with incompatible voice or style settings may be skipped rather than attached to the wrong configuration.
Data Retention
- Local content and audio: Remain on your device until you remove them, clear generated audio, or delete the App. Copies you export remain wherever you chose to save or share them.
- Premium sentence text: Processed to complete speech generation. We do not intentionally store full submitted text in our long-term wallet database.
- Custom scene descriptions: Not stored. The resulting dialogue is cached for up to 24 hours under a one-way hash, only to make retries safe.
- Preset scene dialogues: May be cached on the service for up to 7 days; they contain no personal data.
- Daily dialogue counts: Deleted automatically about 8 days later.
- Dialogue reports: Retained as needed to review how scenes and dialogues are flagged and to improve them.
- Short-lived authentication/session records: Retained only for their configured security lifetime.
- Purchase, wallet, fraud-prevention, and diagnostic records: Retained as needed to operate the service, reconcile purchases, prevent abuse, resolve disputes, and meet legal or accounting requirements, then minimized or deleted when no longer needed.
Your Choices
- Use free System voice if you prefer on-device speech generation.
- Practice the built-in offline scenes or your own text with System voice if you prefer not to use the online services.
- Decline microphone access; reference-audio playback remains available without recording your voice.
- Withdraw your agreement for custom scenes in Settings › Generation; you will be asked again before your next custom scene.
- Clear generated audio or remove articles and dialogues from inside the App.
- Use Settings → Backup & Restore to create or import a user-controlled copy of compatible content and audio.
- Stop using premium voice to prevent future text from being sent for cloud generation.
- Contact us to request deletion of eligible server-side records linked to your App transaction identifiers.
Service Providers
We do not sell or rent personal data. We share limited data only with providers needed to operate the selected feature:
Terms and EULA
Use of the App is also governed by Apple’s standard End User License Agreement:
Changes to This Policy
We may update this policy when the App or its supporting services change. We will update the Effective Date when revised terms are published.
Contact
For privacy questions or eligible deletion requests:
- Email: contact@kuroneko-cmd.dev