Last updated: 8 October 2026
Sublate ("the app", "we", "us") is a real-time translation tool that displays translated subtitles and on-screen text as an overlay. This policy explains what the app accesses, why, and how your information is handled.
Core principle: your media stays on your device; translation is performed online. Sublate has no user accounts and runs no analytics of its own. The audio and screen images it captures are processed only on your device and are never uploaded. To translate, the text extracted from them is sent to an online translation service (Google Translate, and DeepL if you configure it). One further identifier leaves the device: for a YouTube video opened in a browser, its video ID (and, when the video has human-made subtitles in your translation language, that language's code) is sent to YouTube to fetch that video's subtitles in advance. See Sections 1 and 3.
We have no accounts, no backend server of our own, no advertising networks, and no analytics of our own. We do not collect or store your personal data.
Only two kinds of data derived from your use of the app leave your device:
The text you translate, which is sent to the online translation service described in Section 3 to perform the translation.
When you open a YouTube video in a browser, the video ID of that video (for example dQw4w9WgXcQ) is sent to YouTube so that the video's existing subtitles can be downloaded in advance and shown without delay. This is a video identifier, not your screen content and not an account identifier. If the video publishes human-made subtitles in your translation language, that language's code (for example tr) is sent with the request. See Section 3.
(Separately, the app downloads its offline speech-recognition models from public file hosts, and Google ML Kit downloads its on-device translation models from Google's servers — see Section 6. Those are ordinary downloads; nothing of yours is sent with them. The Google ML Kit library bundled in the app may also send its own technical diagnostics to Google — see Section 6.)
The app requests certain Android permissions strictly to provide its features. The raw audio and screen content captured is processed only in memory, on your device, and is discarded — it is never uploaded or stored. Only the text to be translated is sent to the translation service (Section 3).
Accessibility Service — Why it is used: To read the subtitle/caption text shown on screen so it can be translated in real time. What happens to the data: The text is read transiently and sent to the online translation service (§3) to be translated; the result is shown as an overlay. Not stored by us.
Display over other apps (overlay) — Why it is used: To draw the translated text on top of other apps. What happens to the data: No data is read; this only draws on screen.
Screen capture (MediaProjection) — for OCR — Why it is used: When you start screen translation, the app captures screen images to detect text. Requires your explicit consent each session. What happens to the data: Frames are processed on-device to detect text and then discarded — frames are never sent. The detected text is translated on your device when an offline language model is available; otherwise it is sent to the translation service (§3).
App/media audio capture (AudioPlaybackCapture) — for Speech Recognition — Why it is used: To turn audio that is playing from other apps into subtitles using on-device speech recognition. What happens to the data: Audio is processed on-device and discarded — audio is never sent or recorded to a file. The recognized text is translated on your device when an offline language model is available; otherwise it is sent to the translation service (§3).
Microphone permission (RECORD_AUDIO) — Why it is used: Used in two ways. Default: Android groups app/media audio capture (AudioPlaybackCapture) under this permission — the microphone itself is not used. Optional "Microphone Mode": some apps and sites block system audio capture; if you explicitly enable Microphone Mode in its consent dialog, the microphone picks up the sound around your device (the media playing from the speaker — which can also include nearby speech) so that such content can still be subtitled. A persistent notification is always visible while capture is active, and you can switch back to system audio anytime in Speech Recognition settings. What happens to the data: In both modes the audio is processed only in memory, on your device, for speech recognition and immediately discarded — it is never recorded to a file, stored, or uploaded. Only the recognized text is used for translation — translated on your device when an offline language model is available, otherwise sent to the translation service (§3).
Storage (Android 9 and older only) — Why it is used: To save a translated image to your gallery when you choose to. What happens to the data: Only the image you explicitly save is written to your gallery.
Notifications — Why it is used: To show the required ongoing notification while a translation service runs (Android foreground-service requirement). What happens to the data: No data collected.
Translation is performed by online third-party services. The text to be translated (subtitle text, on-screen text, or recognized speech) is sent over your internet connection to Google Translate, and additionally to DeepL if you provide a DeepL API key. These services translate the text and return the result. Their handling of that text is governed by their own privacy policies (see Google's and DeepL's privacy documentation).
Sublate does not store the translated text or send it to any server we operate. On-screen text recognition (OCR) and speech recognition run on your device — only the resulting text is sent for translation, never the raw screen images or audio.
When you watch a YouTube video in a browser, Sublate reads the video's address from the browser's address bar on your device and extracts the video ID from it. That video ID is sent to YouTube (www.youtube.com) to ask for the subtitle tracks the video already publishes, so they can be downloaded ahead of time and displayed without delay. If the video already publishes human-made subtitles in the language you translate into, Sublate also requests that track — that request additionally contains the language code of your translation language (for example tr) — and may display those subtitles instead of a machine translation.
What is sent: the video ID (for example dQw4w9WgXcQ) and, only when such a track exists, the language code of your translation language (for example tr). Not the full URL, not your browsing history, not your screen content, not any account or device identifier.
Who receives it: YouTube — the same service that is already streaming the video to you.
Why: downloading the subtitles in advance makes subtitles appear instantly instead of lagging behind the speech.
YouTube's handling of that request is governed by Google's privacy policy.
Apart from the text sent for translation (above), this is the only case in which anything derived from what you are watching leaves your device. It does not happen for the YouTube app, for other apps, or for other websites.
Sublate is intended for personal use — language learning, accessibility, and understanding foreign-language content. Please respect the copyright and DRM protections of third-party content. The app does not bypass content protection; on protected (secure) screens, capture is blocked by the operating system.
The app is not directed to children under 13 (or the higher minimum age set by the law of your country, for example 16 in some EU countries) and we do not knowingly process children's data. We have no accounts and keep no personal data, so there is nothing for us to collect from anyone of any age.
These are all of the third parties the app can contact:
Google Translate (translate.googleapis.com, clients5.google.com) — used to translate the text you choose to translate, as described in Section 3.
DeepL (api.deepl.com, api-free.deepl.com) — optional; used for translation only if you add your own DeepL API key.
YouTube (www.youtube.com) — receives the video ID of a YouTube video you open in a browser (and, when available, the code of your translation language), so its existing subtitles can be downloaded in advance (Section 3).
Hugging Face (huggingface.co) and GitHub (github.com) — used only to download the offline speech-recognition models to your device. These are plain file downloads; no text, audio or screen content of yours is sent to them.
Google ML Kit (on-device text recognition and on-device translation) — runs on your device; its on-device translation models are downloaded from Google's servers. It is a Google library bundled into the app and it brings Google's own diagnostic transport (com.google.android.datatransport) with it, which may report library-level diagnostics to Google under Google's privacy policy. According to Google's ML Kit data disclosure, this can include device information (manufacturer, model, OS version), app information (package name and version), per-installation identifiers that are not meant to identify you or your device, performance metrics, error codes and, for translation, the configured source and target languages. It does not include the text, images or audio being processed. It is sent encrypted (HTTPS) and is used by Google for diagnostics and usage analytics of the library. We do not receive and cannot read anything it may send.
This version contains no advertising networks and no analytics of our own — we operate no analytics, crash-reporting or attribution SDK, and no advertising ID is requested. If a future version introduces optional ads or subscriptions, this policy will be updated to describe those services before that version is released.
The developer responsible for the app and for the processing described in this policy ("data controller" under the EU GDPR and "veri sorumlusu" under Türkiye's Personal Data Protection Law No. 6698, KVKK) is:
Emirhan Karabulut (Emirsoft), Sivas, Türkiye — sublate.support@gmail.com
Text you choose to translate is sent to Google Translate (or DeepL, if you add your own key) because that is the service you ask the app to perform — GDPR Art. 6(1)(b); KVKK Art. 5(2)(c).
YouTube video ID (and, when available, your translation language code) is sent to YouTube so subtitles can be shown without delay — our legitimate interest in a working, responsive feature, GDPR Art. 6(1)(f); KVKK Art. 5(2)(f). You can avoid it by not using Caption Mode in a browser.
ML Kit diagnostics are sent by the Google library for its reliability — legitimate interest, GDPR Art. 6(1)(f); KVKK Art. 5(2)(f).
Model downloads necessarily reveal your IP address to the file host, as any download does — GDPR Art. 6(1)(b); KVKK Art. 5(2)(c).
Screen images, audio and microphone input are processed only on your device and are never received by us. Microphone Mode additionally requires your explicit choice in its consent dialog.
Text sent for translation: we never receive or store it. Recent translations are kept only in the app's memory to avoid repeating requests and are discarded when the app stops.
Screen images and audio: processed in memory and discarded immediately; never written to a file.
Settings, your glossary and saved layouts stay only on your device until you delete them, clear the app's data or uninstall the app. Your DeepL key is kept in encrypted on-device storage. The app does not use Android cloud backup, so none of this is copied to your Google account.
Downloaded models stay on your device until you delete them or uninstall the app.
Third parties (Google, DeepL, YouTube, Hugging Face, GitHub) keep what they receive under their own policies; we have no access to it.
Depending on where you live, you have rights over your personal data — under the GDPR (Articles 15–22): access, rectification, erasure, restriction, portability and objection; under KVKK (Article 11): to learn whether your data is processed and for what purpose, to request information, to know the recipients in Türkiye or abroad, to request correction or deletion, to object to results of automated analysis, and to claim compensation for unlawful processing.
Because the app has no accounts and we operate no server, we hold no data that can identify you. For data received by Google, DeepL or YouTube, please also contact them directly. Send requests to sublate.support@gmail.com; we will reply within 30 days.
You may also complain to a data protection authority: in Türkiye, the Personal Data Protection Authority (Kişisel Verileri Koruma Kurumu, kvkk.gov.tr); in the EU/EEA, the supervisory authority of the country where you live.
The services listed in Section 6 may process what they receive outside your country — including in the United States (Google LLC) and the European Union (Google Ireland Limited, DeepL SE). These transfers happen because you use a feature that relies on that service, and each provider processes the data under its own terms and privacy policy.
We may update this policy as the app evolves. Before releasing any version that changes what data leaves your device, we update this policy and its "Last updated" date.
For questions about this policy or to exercise your rights, contact: sublate.support@gmail.com