Who is responsible
Vooze Global Kft., 1118 Budapest Rétköz utca 16. földszint 4., Hungary, is the controller for Pinyin Path Firebase accounts, optional cloud AI requests and these pages. Contact info@vooze.app. The original ChatGPT Sites account and its records remain separate unless an explicit migration confirms otherwise; using the same email does not merge accounts.
Accounts are for people aged 16 or older. Registration asks you to confirm this; we do not collect your date of birth. Guest learning stays separate on your device.
What we use and why
Firebase Authentication processes your email, password credentials and account identifier for sign-in and recovery. Cloud Firestore stores signed-in learning progress, answers, review history, writing drafts and settings. A local copy supports learning on your device. Do not put sensitive personal information in practice writing.
If you choose Google sign-in where it is offered, Firebase processes the account identifier, email and any profile name or profile-picture URL supplied by Google for authentication and account management. Which profile fields are provided depends on the information returned by Google. Google sign-in is optional and is available only on platforms where the app offers it.
If you choose Sign in with Apple where offered, Firebase processes the identifier Apple assigns to this app and the email address you share, which may be an Apple private relay address, for authentication and account management.
For signed-in shared reading, we keep a small account-owned retry receipt with the text identifier, saved-state revision and timestamps so a repeated save is not counted twice. These receipts do not contain the reading passage or your practice response. Access is limited to the owning account and authorised backend operations. They are scheduled to expire after 30 days; automatic cleanup may happen later. They are removed with the account’s learner records during deletion.
RevenueCat receives your account identifier, subscription receipts or purchase tokens, entitlement status and technical connection information to verify and restore membership. This may happen when you sign in even while new purchases are unavailable. We do not send your email as a RevenueCat customer attribute. Apple or Google handles payment; we do not receive your full payment-card details.
Our billing backend temporarily queues a limited event summary with account and subscription identifiers so membership updates can be retried safely. It does not retain the full billing webhook payload. The queued identifiers are removed after successful processing; unresolved event summaries are scheduled to expire within 30 days. Aggregated billing status and legally required records follow the retention rules below.
Support messages contain the information you choose to send. We use account and learning data to provide the service you request (performance of our contract), security records for our legitimate interest in preventing abuse and keeping accounts safe, and billing records where legal obligations require them. Permission-based microphone use and reminders are optional.
Your voice, reminders and this website
Local speaking practice keeps its clips on your device. The app attempts to remove them when replaced or closed; interrupted cleanup or operating-system caches may leave files behind. Cloud transcription is a separate optional action described below. You can revoke microphone permission in device settings. Reminders are scheduled locally and can be turned off.
The app has no advertising or cross-app tracking. We do not enable a behavioral analytics SDK. RevenueCat processes purchase history for membership checks and subscription analytics, including while checkout is unavailable. The static policy pages set no analytics, advertising cookies or external fonts. The learning app uses device storage for preferences, progress and sign-in; hosting providers process connection and security information such as IP addresses.
Starting with iOS version 1.3.1, Firebase Crashlytics automatically processes crash stack traces, app and device details, OS version, SDK-generated installation/session identifiers and limited session or transport diagnostics to help us detect and fix crashes. We use this for our legitimate interest in a reliable service. We do not attach your account ID, writing or recordings as custom crash fields or logs, and do not add Google Analytics. There is currently no in-app crash-reporting switch. Firebase states that crash traces and associated identifiers are kept for 90 days before removal from live and backup systems begins; account deletion does not instantly remove these separate diagnostics. See Firebase privacy information. This native iOS integration is not installed in the Android app or website.
Optional AI and your choice
Where enabled, AI coaching sends the learning text you choose to OpenAI through our authenticated server; narration sends selected text and returns an AI-generated voice; cloud transcription sends the short recording you choose and returns a transcript. We ask for your permission before cloud AI use. You can decline or revoke it in the app, stop recording, and continue non-AI learning. Microphone permission alone does not authorise cloud AI. Do not submit sensitive personal information. A transcript is not a pronunciation score, and the narration is synthetic, not a real teacher’s recording.
For optional AI practice in the mobile app, we use your consent. Our gateway processes request audio and text transiently and does not create a persistent AI prompt, audio or transcript history. Learning work you separately save follows the ordinary learning-record rules. Account-linked usage counts and limited request metadata are kept to prevent abuse and enforce the displayed allowance, on our legitimate interest in service security. They are scheduled to expire after 30 days; automatic cleanup may be later. Account deletion removes this account’s AI usage records. Revoking consent stops future requests but does not undo completed processing or override lawful provider retention.
If you choose to report an AI coaching response and confirm the report, Pinyin Path stores the response excerpt shown in the report form, your selected reason, optional note, account identifier and submission time to investigate the issue and improve coaching safety and quality. The excerpt may repeat words from your practice. Your original draft or recording is not attached separately, and the report is not sent to OpenAI or RevenueCat. Reports are accessible only to authorised backend and support operations. Reports are scheduled to expire after 30 days; automatic cleanup may happen later. Completing account deletion removes the reports and their account-owned anti-abuse counter.
Near-live listening has an additional permission choice. When you press Start, short microphone segments are automatically sent to OpenAI through our authenticated server for transcript updates. Capture stops after at most 30 seconds, with up to six segments; short capture or update pauses can occur. Stop or withdrawing this permission prevents further capture and uploads, but processing of already-sent audio may finish. The transcript is held on the screen, not saved as a learning record. Server-only session receipts contain identifiers, status and byte/duration counts, without audio or transcripts. They are scheduled to expire after 24 hours; cleanup may happen later.
In the mobile app, natural AI narration has a separate choice for the current session. We may cache synthetic audio of our authored course text for reuse; this shared course audio contains no learner draft or recording. Private text narration is not cached. For contextual speaking help, you first choose one recording for transcription; a further action sends the recognized words and current exercise for text feedback. Those suggestions do not evaluate the sound of your pronunciation. Closing help or changing context dismisses late results but cannot undo processing already sent to the provider.
The web reading library uses separate account-owned storage hosted on Cloudflare. When web generation is available, it stores the requested topic and settings, learned-vocabulary context, generated readings and novel outlines or chapters, request and status records, and generated narration audio so work can be reopened and audio replayed. Web speaking help can retain the recognized pinyin, meaning and feedback with its request record; this web route does not save the microphone recording. Pausing generation does not erase saved web records or cached narration. These records do not follow the native gateway’s 30-day counter expiry; they remain while the account is active, until account deletion or a support-assisted erasure request.
Web background writing uses stored OpenAI responses while results are prepared. After processing the result, we request deletion of the stored provider response. If cleanup is incomplete, its identifier remains so account cleanup can try again. The provider’s applicable retention rules still apply.
Providers and international processing
Google Firebase operates authentication, cloud storage, backend functions and website hosting for us. Our learner database is configured in Frankfurt; authentication and other processing can occur outside the EEA, including the US. Firebase privacy information describes Google’s processing and transfer arrangements.
RevenueCat processes subscription data in the US. Its published processing terms include standard contractual clauses for relevant international transfers; see also its privacy policy. These providers and authorised service operators access data as needed to deliver or secure the service. The service is not end-to-end encrypted.
OpenAI processes the selected AI inputs to deliver the requested result. Its API data controls explain endpoint-dependent retention, including standard abuse-monitoring logs that may contain request content and are normally retained for up to 30 days, with stated exceptions. We do not promise zero retention or EEA-only processing. Requests may be processed outside the EEA. Ask us about the applicable processing and transfer arrangements; your permission does not replace required provider safeguards.
Retention and account deletion
We keep account learning data while your account is active, until deletion or an erasure request. In Profile, open your account and choose Delete account. Completed deletion removes Firebase authentication, learner cloud records, account-linked AI usage records, the RevenueCat customer, admin roles/previews and the signed-in copy on this device. Your exports, separate guest progress and the original Sites web account are not automatically erased. Contact us about legacy web data.
A failed deletion keeps a protective record tied to your account identifier and pauses new uploads until retry completes. After completion, a minimal identifier/status receipt is scheduled for expiry after 30 days; automated cleanup can occur later. Provider backups and operational logs follow their own retention processes rather than disappearing instantly.
Limited administrator configuration and security audit records, including the administrator’s identifier, remain outside learner deletion for security and accountability, for as long as needed to investigate events or demonstrate configuration decisions. Statutory billing records may also need to be retained. Contact us about review or erasure of these records.
Deleting the app or account does not cancel store subscriptions. Manage billing separately. If in-app deletion is unavailable, request deletion at info@vooze.app.
Web-library records belong to the sign-in identity that created them. For export or erasure, use the available account controls or contact info@vooze.app. A Firebase account export includes its same-identifier private library and request records; generated audio files are not included in that download. Completed Firebase account deletion also removes web-library records and generated audio under that same identifier and requests cleanup of stored provider responses. A legacy ChatGPT Sites identity remains separate; contact us about its export or erasure. Matching email addresses do not merge these identities.
Your choices and rights
You may request access, correction, erasure, restriction, portability or object where applicable. Email info@vooze.app; never send your password. We may verify identity. We normally respond within one month, explaining any lawful extension. See the EDPB rights guide.
You may complain to Hungary’s NAIH or your relevant data-protection authority. Learning recommendations use your activity to suggest practice; we do not make automated decisions with legal or similarly significant effects. Material changes to this notice will be explained in the app or on these pages before new processing takes effect.