SuperTalk privacy policy
Updated November 3, 2026. Milly Works Inc. provides SuperTalk for text, camera and voice translation.
Your permission for AI processing
AI sharing is off until you explicitly choose Allow in the app’s AI data-sharing disclosure. Not now leaves it off. You can withdraw permission in Profile → Online translation → Use offline only. This blocks new AI requests and stops active online sessions; it cannot recall data already processed. Downloaded offline features remain available. Microphone permission and keyboard Full Access do not replace AI-sharing consent.
Translation
Online translation sends the text you submit, recognized camera words, or microphone audio to our server and to our AI partners for translation, transcription, pronunciation and speech. They are named under Retention and providers below. Camera text recognition happens on your phone: camera images and selected translation photos are not uploaded for translation. Downloaded offline translation runs on your phone.
During a joined conversation, your display name, speech transcript and translations are shown to the other participant. You decide when to start and stop recording. Lock-screen recording continues only during a conversation you start.
Accounts and saved content
Signing in is optional. Google or Apple provides your account identifier, email and, when available, name. We use Supabase for account authentication and storage, and PowerSync to synchronize profile preferences, keyboard settings, phrases you choose to save and learning progress. A profile photo you choose is uploaded to private account storage. Guest settings and saved phrases stay on your phone until you sign in.
Keyboard
With online translation enabled, text composed in the optional SuperTalk keyboard’s translation box is sent for translation after typing pauses. Audio you choose to dictate is sent for online processing. Password and other direct-input fields bypass translation. You can disable online translation in Profile or remove the keyboard in phone settings.
Screen translation and live subtitles
When you start screen sharing for translation, text recognized on your screen is sent to our server and OpenAI for translation. Recognition runs on your device; screen images are not uploaded. When you start live subtitles, audio playing in other apps is sent to our server and Soniox for transcription and translation. You control when screen sharing and subtitles start and stop. These features require the same prior AI-sharing permission.
AI tutoring
When you allow AI sharing and use tutoring, we send your submitted text or audio, name if provided, languages, recent conversation history, saved phrases, learning goals, progress and tutor memories to our server and OpenAI to generate personalized lessons and speech. Selected tutor voice modes use xAI for audio and conversation context. Pronunciation assessment uses Microsoft Azure Speech for recorded audio and reference text. Tutor messages, goals and memories may be saved with your account to continue your learning.
Usage analytics
To learn which features people use and where the app gets stuck, the app and our server send usage events to PostHog: for example that setup finished, a conversation started and how long it lasted, a translation or purchase happened, or which screen opened. Events include your language choices, app version, phone model, approximate location derived from your IP address and, when you are signed in, your account identifier. They never include what you type, say or photograph, translations, names or email addresses. We do not use them for advertising or share them for tracking.
Retention and providers
We keep saved account data until you remove it or delete your account. Active conversation state is held while the conversation runs; our application does not archive conversation recordings or log translation content. Cloudflare hosts our translation server and may process connection and operational information such as IP addresses.
Our AI partners are OpenAI (translation, pronunciation and speech) and Soniox (live speech recognition and live translation of microphone audio). Soniox processes live audio as it arrives, does not store it, and does not use it to train its models; see Soniox privacy policy.
OpenAI may retain request content for abuse monitoring under its service policies, normally up to 30 days unless another retention setting applies. We disable stored text Responses requests. API content is not used for model training by default. Provider and infrastructure retention may outlast account deletion. See OpenAI data controls.
Additional AI processors used by tutoring are xAI (voice conversation) and Microsoft Azure Speech (pronunciation assessment). Their processing and retention are governed by their service terms and your feature use: see xAI enterprise terms and Microsoft Speech data privacy. We require third parties processing personal data for SuperTalk to provide protection equal to or greater than the protections described in this policy, including appropriate security and limits on use and disclosure.
Your choices
You can deny microphone, camera or photo access, use downloaded offline features, remove saved phrases, sign out, or delete your account. Account deletion removes the account, saved cloud phrases, progress, settings and profile photo. Copies already downloaded on another phone may remain there until you remove the app.
We do not use advertising SDKs, sell personal information or track you across other companies’ apps. SuperTalk is a general audience service and is not directed at children. Do not rely on AI translations for emergency, medical or legal decisions.
Contact
For access, correction, deletion or other privacy requests, contact hello@getsupertalk.com. We will update this page when our practices change.