Privacy Policy · Klee Translate

01 · Who we are and how to reach us

Klee Translate is developed, operated and maintained by 上海市长宁区湖歌译翻译服务工作室, an individual business (个体工商户) registered in Shanghai, mainland China. “We” in this policy means that business, which is also the controller of personal data in this service; “you” means anyone visiting the website or using the extension and the services that go with it. We are responsible for account management, translation processing, subscription billing and user support within this service.

The operator is located in mainland China, so your account and translation data are controlled by an operator inside China. The service is not offered to users in mainland China; that is a limit on where the service is sold and used, which is a separate matter from where the operator is. Which recipients process data, and where, is set out in section 05.

Support, privacy requests and account deletion all go to support@kleetranslate.com. Not offering the service in mainland China is a limit on service availability; it does not affect the right of anyone who has already given us personal information to request access, correction or deletion under the law that applies to them.

02 · What data we process

Different features need different data. Web page and selected text are submitted when you start a translation yourself, or when a page matches the automatic translation you switched on. On a streaming page the extension captures the current subtitles; with automatic subtitle import enabled, the subtitles and episode details are sent to the backend to identify the episode, estimate usage and prepare for translation — this can happen before you confirm a paid translation. Subtitle diagnostics may also run on that page.

Account and authentication: Username, email address, account identifier, email verification status, sign-in tokens and authentication results. When you register or sign in, the password you type is passed by our backend to the identity service; the application backend does not store passwords in the clear and does not write them into application logs. This has nothing to do with passwords for other sites you visit.
Web page and selection translation: The page text you select or ask to have translated, the context needed around a selection, the page's domain, title, description and main heading, plus the target language and the model you chose. The web cache stores text fingerprints and translations; it does not keep the page's source text in its cache fields, although the source text is still processed and sent to the model provider when the translation runs.
Streaming subtitle translation: The current episode's subtitle text and timings, subtitle source information, the platform's video identifier, the show title, season and episode numbers, the chosen model and target language, and the resulting translations, episode summaries, translation jobs and progress. We do not upload video or audio for translation.
Document translation (PDFs and images): PDFs and images you open in the translation tools on our website (www.kleetranslate.com/tools/) are parsed in your browser (images are turned, on this device, into a PDF containing only those images); the original file, its file name, the parsed text and the translations are stored on this device and are not uploaded to our servers. When you request an estimate or start a translation, the page uploads only the paragraph text of the pages you selected, their page numbers and paragraph types, plus a few preceding paragraphs and headings as translation context and the document title used as a hint for the model (taken from the first heading). It never uploads the PDF or image file itself or its file name (for the page images sent when you choose enhanced recognition, see section 05). Our servers also keep task status, page counts and credit charges, without any text or file names. When you export a PDF that keeps the original layout, or translated images, the page downloads the font files needed for the target language from our website; these requests contain no document content.
Preferences and personal entries: Subtitle display and translation settings, the list of sites or shows set to translate automatically, personal terminology and name rules, and any corrections or reference packs you provide, used to keep your preferences, keep names consistent and restore your state.
Subscription, payment and credits: Plan, billing period, credits used and rolled over, order numbers, amounts, payment transaction identifiers, payment status, refund requests and how they were handled. Web page charges may include the page's domain and title. Payment happens at the payment provider's own checkout; we do not receive the payment password or full card details you enter there.
Email notices and sending records: To select recipients, use the appropriate language and prepare subscription-expiry reminders, we process the account email, account identifier, username, interface language (the one the extension reports; until it has reported one, the language preference your browser declares in its requests), plan and expiry time. Application sending records contain the recipient email, associated account, message type, language, subject, sending time, status, retry count and error information, plus the delivery outcome Cloudflare reports back (delivered, deferred, bounced, or marked as spam by the recipient), a summary of the receiving server's response and the message ID. We use them to prevent duplicate messages, investigate delivery problems and handle support requests; individual delivery records do not store the full message body; campaign records separately retain the subject, body, category and language composed by an administrator. Test messages sent by an administrator also record the test recipient address they provide.
Referral activity: When you register through an invitation link, we record its code, both account identifiers, a normalized email for duplicate-claim prevention, and verification, first-order and reward status. We use this to attribute referrals, grant credits, prevent duplicate claims and handle refunds. Inviters see aggregate progress and their own rewards, not friends’ names, emails or translated content. These records follow the retention rules for account, transaction and support records.
Operations, security and support: Request times, IP address, browser or client information, errors and timings, model usage, and whatever you send to the support address. Subtitle diagnostics may also include the suspected subtitle request URL, its opening fragment, player elements and an outline of the subtitle tracks, used to find out why capture failed; this information can reveal what you are watching.

The service does not collect your full browsing history, bookmarks or the contents of unrelated tabs, does not read sign-in credentials for other sites, and does not use translation features to collect what you type into password fields, form inputs or editable areas. Text you select yourself may still contain personal information, so only submit content you are entitled to process.

03 · Caching is per user

Subtitle and web translations are cached per account; there is no shared cross-user translation cache. You can reuse results already in your own account when the model, target language and content match; two users translating the same page or the same episode never read, continue or overwrite each other's translations or job progress.

Personal entries, name preferences and usage records are not made available as another user's personal data.

One layer is shared per show: the reference pack. While an episode is translated, the system automatically extracts proper nouns and their translations (character names, place names, in-universe terms) along with style points that apply to the whole episode, and stores them as shared entries for that show and target language; other users translating the same show draw on them so that names stay consistent. A reference pack contains only entries of that kind — no account identifiers, no viewing records, and no note of who translated which episode — so nobody can use it to read your translations, job progress or personal entries. Automatically extracted style points do not carry their source-text evidence when they are used across accounts.

This isolation is an access and reuse rule on the server. It is not end-to-end encryption, and it does not mean data stays only on your device. To perform a translation, the backend and the relevant model providers still process the necessary text; authorised operations staff may access data only as far as needed to provide support, investigate security issues or meet legal obligations.

04 · Purposes and legal bases

To provide what you asked for: registration and verification, producing translations, keeping your personal cache and preferences, fulfilling subscriptions, settling credits and handling support.
To keep the service safe: preventing account theft, repeated trial claims, abuse and fraud, diagnosing failures and accounting for service costs. We process those records only as far as necessary.
To meet legal obligations: keeping the transaction records required by applicable law and handling lawful requests and disputes.

Where the applicable law requires a legal basis to be stated: providing accounts, translation and purchases generally rests on entering into or performing our contract with you; necessary security and troubleshooting rest on legitimate interests, weighed against your rights; statutory retention rests on legal obligations. Optional processing that requires separate consent will be explained and asked for separately, you can withdraw it, and withdrawal does not affect processing that was lawful before.

Service email and optional subscriptions

We may send necessary notices to your account email about account security, service availability, the status of a purchased service, or significant changes to our Privacy Policy or Terms of Service. These messages are limited to facts relevant to you and any necessary action, without product promotions or special offers. Depending on the purpose of the notice, processing rests on performing our service contract, meeting a legal obligation, or our legitimate interest in keeping accounts and the service working, weighed against your rights.

Product updates, offers and promotions intended to introduce features, recommend products or encourage purchases or renewals are optional marketing. In the extension settings → My account → Email preferences, you can separately choose Product updates and Offers. Both are off by default; we send a category only after you actively select and save it. Accepting this policy, creating an account, signing in with Google or purchasing a plan does not constitute consent to marketing email.

Optional email relies on your consent. We retain the category, email, consent-text version, time, source and withdrawal records. A changed email requires a fresh choice. You can turn off a category in Email preferences or use the free unsubscribe link in each optional message without signing in. Opening a link does not unsubscribe you until you confirm; email clients supporting one-click unsubscribe can submit the request directly. If you mark an optional message as spam in your mailbox, we treat that as withdrawing consent for that category, stop sending it and note the source in the withdrawal record. Withdrawal does not affect purchased services or necessary security and policy notices. Messages already submitted to the delivery service may still arrive. You can also contact support to request access, deletion or to object to processing.

We do not sell or rent personal information, we do not use translation content, browsing activity or account data for targeted advertising or cross-site profiling, and we do not use your translation content to train our own general models. In line with the Limited Use requirements of the Chrome Web Store user data policy, we use and transfer extension data only for the user-facing features described here and for permitted security and legal purposes; human review is limited to permitted cases such as support you explicitly agreed to, necessary security investigations or legal requirements.

05 · Service providers and data transfers

Delivering the service requires the following recipients to process the necessary data:

Model providers: the model calls required by your model choice and by the auxiliary processing a feature explicitly describes. Our server sends the text to be translated, the necessary context and glossary information to that model's API and receives the translation; we do not send your sign-in password or payment credentials to model providers for translation. For document translation we send only the paragraph text of the selected pages and the context described above, never the PDF file itself or its file name. The available providers are DeepSeek, OpenAI, Anthropic, Moonshot AI, Google and xAI, as listed in the product at the time; of these, DeepSeek and Moonshot AI process the text they receive inside China. You choose the model in the extension, and thereby which company handles a given translation. Each API has its own retention and use rules, so per-user isolation on our side must not be read as a guarantee of zero retention by a third party.
Baidu enhanced document recognition (optional): only after you choose enhanced recognition for a document, images of the pages requiring recognition (for an image document, the content of the images you imported) are relayed through our server to the PaddleOCR-VL document parsing service on Baidu Intelligent Cloud to obtain text, formulas and layout coordinates. The original file itself, real filename and password are not sent. If you do not choose enhanced recognition, text in scanned pages and images is not recognised or translated, and the original images are kept. Our server processes page images and recognized text only during requests, without persisting or logging them. Baidu Intelligent Cloud processing and retention follow its service terms. Official result links last 30 days; this is not a deletion deadline for all copies. We cannot promise zero third-party retention.
Infrastructure, identity and email services: hosting, databases, network protection, authentication and the services needed to send verification or password-reset email, each processing the data their role requires.
Cloudflare email delivery: subscription-status notices, necessary notices and opted-in product updates and offers are sent through Cloudflare Email Sending. We provide the recipient email, sender and reply-to addresses, subject and message body for delivery; the body may contain the username, plan and expiry time. Cloudflare also processes delivery status, bounces, spam complaints and related technical records, and reports each message's delivery outcome back to us; we use this for delivery, troubleshooting and honouring your unsubscribe choices. We do not submit subtitles, translation content or sign-in passwords for sending these messages. See Cloudflare’s Privacy Policy for the provider’s rules. Applicable cross-border transfer safeguards must also be in place when using an email provider.
Payment: this service takes payment only through Creem. Card and other payment details you enter at the checkout are handled directly by Creem and its payment partners; we cannot see them and do not store them.
Creem (payment provider): purchases settled through Creem have Creem as the Merchant of Record handling payment, the related taxes and billing. To complete a purchase and grant the entitlement, we pass the account email, product, order and amount required by the actual transaction, and receive the necessary customer, country or region, order and payment status; we do not send subtitles or the text to be translated for the purpose of taking payment. Data you provide at the checkout, and Creem's own processing, are covered by their privacy policy.
Website resources: the fonts on the portal pages are hosted by us and no request is sent to a third-party font service. Once you follow a link to another website, that website's privacy policy applies.

Your data is processed inside China. The operator is in mainland China, and DeepSeek and Moonshot AI also process the text they receive inside China; OpenAI, Anthropic, Google, xAI and the infrastructure providers may process data in the United States or elsewhere. China has no adequacy decision from the European Commission, so for users in the European Economic Area or the United Kingdom this is a transfer to a third country, which requires valid transfer safeguards such as standard contractual clauses together with an assessment of the circumstances; this policy is not a substitute for the safeguards or the separate consent the law may require. The exact location of servers and databases depends on the deployment; write to the support address if you need to know the model you are using, where the infrastructure is, or what transfer arrangements apply.

Beyond the disclosures needed for the functions above, we disclose data only where the law requires it, where it is necessary for security and the protection of rights, or where you have separately and explicitly authorised it. Personal caches are never made public to other users.

06 · How long data is kept

In your browser: Preferences, sign-in tokens and the local state the extension needs are kept in extension storage. Signing out removes the local session; uninstalling the extension clears its local storage but does not automatically delete your server-side account or transaction records.
Subtitles, translations, jobs and personal settings: Kept so that you can carry on using them, until you delete the data, clear your usage traces or ask for the account to be deleted — except for material that applicable law requires us to keep or that is needed to resolve a dispute.
Web translation cache: Isolated per account, currently kept for at most 180 days, and possibly evicted earlier when capacity requires it; after that it is no longer used for cache hits and is removed by the cleanup job. Translating again may create a new cache entry.
Temporary document translation content: Paragraph text, the document title and translations not yet downloaded are kept only to carry out the translation: incomplete uploads for at most 1 hour; uploaded text is deleted when that translation task ends, and at the latest 24 hours after the first upload; a translation is deleted as soon as your extension downloads it and confirms it has been saved locally, and translations never downloaded are deleted after at most 7 days, after which they cannot be retrieved. This content is kept in dedicated database tables that are not replicated to standby instances. The disk holding the database is snapshotted every hour for disaster recovery and only the 5 most recent snapshots are kept, so temporary content still being processed when a snapshot is taken stays in that snapshot and is deleted with it about 5 hours later at the latest. Snapshots are used only to restore the service, access to them is restricted, and they are not used for translation or any other purpose; if we ever restore from a snapshot, we re-apply deletion requests already received and the periods above. Task status, page counts and credit charges, which contain no text or file names, follow the account and transaction rules; status records of finished tasks are removed after 7 days.
Enhanced recognition progress and results: Request identifiers and completed recognition results are stored locally per account to resume interrupted work without submitting the same recognition again. Deleting the associated document, clearing browser data or uninstalling the extension removes them. Progress for an import that never reached the document list can be removed by clearing extension data. Our server stores only task and usage metadata, without content. Enhanced recognition charges credits per page at the price you confirm for each document; failed pages are refunded and translation is billed separately. Unfinished tasks expire after 15 minutes; results remain retrievable for at most 7 days after the first successful retrieval (or less if Baidu Intelligent Cloud expires its result earlier), and vendor task identifiers and image hashes are cleared after the extension confirms local storage. Content-free task metadata is removed 7 days after retrieval eligibility ends. Clearing usage traces or closing an account removes server-side recognition tasks; credit entries remain under the transaction-record rules. These actions do not delete copies already processed by Baidu.
Documents and translations on your device: The PDFs and images you open in the translation tools on our website, their file names, the parsed text and the translations are kept in this device's browser storage for our website (not in the extension), separately for each signed-in account, and are not synced to other devices, until you delete them from the document list or clear this website's browser data. Passwords you enter for a document (its open password or the author's permissions password) are likewise stored only on this device, to reopen and export that document, and are never uploaded. Once removed we cannot restore them, and translating again is charged as a new translation task; export anything you want to keep.
Application diagnostics and cost records: Application error logs are cleared on a 24-hour window, timing metrics on 7 days and model cost records on 90 days. These windows are not the same as the retention of payment, identity, email or infrastructure logs.
Application email sending records: Individual delivery records, campaign bodies and preference-change history are periodically cleared on a 180-day window. Personal unsubscribe links expire after 180 days. Current preferences and their consent version, time and associated email are kept with the account to enforce your choice until you change it or delete the account; expiry of change history does not resubscribe you. Closing an account deletes its preferences, consent records, unsubscribe tokens and individual delivery records; clearing translation usage traces alone does not delete them. Cloudflare’s delivery logs and any retained message content are governed by its service settings, terms and applicable law, and do not necessarily follow our application’s cleanup period. Deleting data from our server also does not remove copies already delivered to a recipient’s mailbox.
Account, transaction and support records: Kept only as long as necessary to provide the account service, reconcile orders, handle refunds and disputes, prevent fraud and meet applicable statutory retention duties. When we handle a deletion request we will say which categories still have to be kept and why; once they are no longer necessary they are deleted or stripped of identifiable links.

Deletion does not mean rewriting every historical backup on the spot. Residual data in backups is kept only for recovery and security, is access-restricted, and is cleared as backups rotate; it is not used for day-to-day translation. If a backup is restored, deletion requests already accepted are applied again.

07 · Managing your data

In the extension's settings you can change preferences and personal entries, turn off automatic translation, or revoke site access in your browser.
“Delete my usage traces” clears the subtitles, translations, translation jobs, viewing records, web translation cache, reference packs, personal entries and preferences in your account. It also removes the page title and domain from web charge records, and removes the direct link between cost records and your account.
Deleting a document from the document list in the extension removes the original and its translations on this device, together with any related temporary text and translations not yet downloaded on our servers. “Delete my usage traces” and account deletion also remove temporary document translation content and task records from our servers; credit charges are kept as transaction records. Documents stored locally on devices that are offline are not deleted remotely.
Clearing usage traces is not the same as closing your account or getting a refund. The credit amounts needed to reconcile your balance, along with subscription, order and trial-eligibility records, may still be kept. Uninstalling the extension does not close your account either.
To access, correct, obtain a copy of or delete server-side data (including caches), to close your account, or to find out why something is kept, write to support@kleetranslate.com. To prevent mistaken deletion or disclosure we will reasonably verify that the account is yours; we will never ask for your password or a payment password.

Depending on the law where you are, you may also have rights to restrict or object to processing, to data portability, to withdraw consent and to complain to your local data protection authority. We reply within the statutory period; under the GDPR that is normally one month, and where an extension is justified we will explain why within the original period. Where the law does not allow us to meet a request in full, we will state the basis and the remedies available, and we will not treat you unfairly for exercising your rights.

08 · Local storage and browser permissions

The service uses the local storage and session mechanisms it needs and no advertising or tracking cookies; the authentication pages may use cookies required to keep a session secure. Portal sign-in credentials are not written to page localStorage for later use. When you sign in from the extension, they are handed to the authorised extension, which keeps the session. When you sign in to the translation tools on our website, the server sets an HttpOnly cookie in your browser used only to renew the sign-in (kl_web; page scripts cannot read it, and it is sent only to the renew and sign-out endpoints); it lasts as long as the sign-in session and is cleared when you sign out. The access token is kept only in the page's memory and is discarded when you refresh or close the page. The website tools also keep the interface language and target language you choose in this website's local browser storage.

IndexedDB (the translation tools on our website): keeps the PDFs and images you open in the website tools, their file names, the parsed text, translations, download receipts and any passwords you enter for them, separately for each signed-in account, and does not upload them; font files used to export PDFs and translated images are kept in this website's browser cache. This feature is not part of the extension, which adds no browser permissions for it.
storage: keeps extension preferences, sign-in state and the local records needed.
activeTab: works out which page you are on and what is available when you click the extension.
webRequest and named site permissions: capture subtitle request information on supported video platforms, or read the current subtitles; not used to read payment credentials.
Page access: reads the relevant text and inserts translations on pages where you started a translation or configured automatic translation. The browser may describe a broad range of site access; that does not mean we continuously upload the content of every page.

09 · Security and minors

We protect data with HTTPS transport, account authentication and access controls. Model API keys are kept on the server and never shipped inside the extension. No system can guarantee absolute security; if you find your account has been taken over or suspect a data breach, contact support promptly and we will investigate and act under the applicable law, notifying affected users where required.

The service is not directed to anyone under 16. Some EU member states set a lower age of consent for information society services (13 at the lowest), and in those places the local rule applies; in no case is the service directed to children under 13. Minors elsewhere must meet their local age requirements for online services and the processing of personal information, and obtain guardian permission where the law requires it. If you believe we are processing a child's personal information without a legal basis, contact support so that we can verify it and delete the data or take other necessary steps.

10 · Changes to this policy

This policy will change as features and data handling change, and each version carries its date. Significant changes to the purposes of processing, the recipients of data or your rights will be announced in advance by an appropriate means — the website, a notice in the extension or the email address on your account — and where the law requires fresh consent, we will ask for it separately. We will not use a policy update to retroactively widen the permissions covering data we already hold.

For questions about data handling or a privacy request, write to support@kleetranslate.com. Purchase and usage rules are in the Terms of Service.