FreeVoice Privacy Policy

Effective date: 20 October 2026 · Updated: 20 October 2026

FreeVoice ("the app", "we", "us") is a voice-to-text dictation app: you speak, and the transcribed text is placed into the supported text field where you started dictating. This policy explains what data the app handles, why, and who processes it. We have tried to collect as little as the product needs.

Who is responsible

FreeVoice is operated by Slava Marchenko, an individual developer publishing as SlavaMarchenko. For any privacy question or request, contact rota2000.2@gmail.com.

What we collect, and why

Account information. When you sign in with Google, we receive a stable user identifier, email address, and the display name and profile-photo URL made available by Google. The name and photo are used to display your account. The identifier is the key that links you to your subscription and your free allowance across devices and reinstalls while the account exists. Deleting the account removes that identity and synced account data. Usage counters remain temporarily under the old opaque identifier until their automatic expiry (the longest configured metering window plus two days; currently 32 days), and cannot attach to a later sign-in. A later sign-in creates a new account with a fresh allowance. We do not retain a device fingerprint or deleted-identity record to link the accounts. We do not receive your Google password.

Your voice (audio). While you are actively dictating, the app records your speech and streams it through our transcription relay to a speech-to-text provider. Our relay keeps the audio only transiently in memory for the duration of a single dictation (so it can be re-sent if the connection drops) and does not persist it. It normally streams the audio to Gemini 3.5 Transcribe Live on the paid Gemini API. FreeVoice does not upload recording files to Gemini or enable its optional saved session history. Google does not use paid API prompts or responses to improve its products, but may log them for a limited period for abuse prevention, safety, and required legal disclosures. If Gemini fails, the ElevenLabs fallback may receive the same audio under its standard request-retention settings; we have opted out of using new ElevenLabs data to improve its models. FreeVoice itself does not train speech models.

Transcribed text. The text produced from your speech is returned through our relay to your device and inserted into the field you are typing in. We do not persist a copy in FreeVoice databases. The paid Gemini API and ElevenLabs fallback handle output under the retention terms described above.

Your Dictionary. The words you add to your personal Dictionary are saved to your account so they sync across your devices. They are sent to our relay with each dictation and forwarded to Gemini as custom vocabulary to improve recognition. They are not forwarded to the ElevenLabs fallback.

Usage for your allowance. We measure the amount of dictation audio you use to enforce the allowance for your plan. Production counters are grouped into UTC calendar days: Free uses the current UTC day and the previous six UTC days; paid tiers use the current UTC day and the previous 29 UTC days. This is a duration measurement, not a recording.

Subscription and purchase status. RevenueCat manages entitlement and purchase status. Its SDK can create an anonymous installation profile before sign-in and receives your FreeVoice user identifier when you sign in, including on the free plan. It processes app/device metadata and any associated purchase and subscription history. If you subscribe, Google Play handles payment. We receive your plan and subscription status, not your payment card or full billing details.

Information used locally on your device. With Accessibility enabled, FreeVoice inspects accessibility nodes, window and foreground-app information, and the active editable field to find the insertion target and hide the Bubble where appropriate. It may read existing text in that field to preserve it while inserting your dictation. Clipboard-based insertion can temporarily read and replace clipboard content, then restore it; a manual-paste fallback leaves the transcript on the clipboard until you replace or clear it. Hidden Apps lists launchable apps so you can choose where the Bubble is hidden. Existing field text and clipboard contents are handled locally and are not uploaded. The launchable-app list and foreground-app names are not sent as product analytics. Support receives only the separately disclosed information you choose to submit.

Relay and service metadata. Our hosting and network providers process connection information such as IP addresses, request times, error status, and protocol metadata to deliver and secure the service. Our relay checks Firebase authentication and App Check tokens and uses your user identifier to check entitlement and meter usage. Operational logs do not intentionally include audio, transcript, clipboard content, or authentication tokens. The infrastructure providers are listed below.

App and device integrity (attestation). Each dictation includes an attestation token (Google Play Integrity via Firebase App Check) that proves the request comes from a genuine, unmodified copy of the app. This is used only to block abuse and fraud.

Diagnostics. We collect crash reports and error diagnostics (via Firebase Crashlytics) to find and fix problems. These contain technical information about the failure and the device, including insertion method, cursor positions, timing, and success/failure flags. They do not intentionally include dictated words, existing field text, or clipboard contents.

Product analytics. Android releases use Firebase Analytics to understand setup completion, dictation reliability, and subscription-screen performance. Collection starts automatically when the app starts. It includes basic app/device and interaction information plus event names such as sign-in, permission completion, onboarding actions, bubble shown, dictation started/inserted/failed, paywall opened, checkout state, and restore state. Our custom events contain no audio, transcript, text-field or clipboard content, app/package name, email, price, or product ID. Advertising-ID collection and personalized-ad signals are disabled. Google Play also supplies purchase and subscription events to Analytics, including product identifiers, amounts, currencies, renewals, cancellations, and refunds. These store events do not contain your dictation audio or text, or payment-card details.

Support notifications. Firebase Cloud Messaging assigns this app installation a Firebase Installation ID (FID). We register that ID under your signed-in account only to notify this installation about a support update. The notification contains a generic event type and ticket identifier, not the supporter's message or other ticket content. The app attempts to remove its registration when you sign out; the server also removes invalid registrations and all registrations for a deleted account. A stale registration cannot display account-specific content after sign-out because every push must match the currently signed-in account before the app shows it.

AI Support. If you choose Ask AI and accept its first-use disclosure, your question, recent AI Support conversation, requested response language, a curated public FreeVoice usage manual, and a small allowlisted app-state snapshot are sent to Google's Gemini API to answer setup and app-use questions. The snapshot may include app and Android versions, whether the Bubble and required permissions are enabled, dictation mode, Dictionary and hidden-app counts, entitlement kind, and remaining allowance. It excludes your identity, device model, app names, dictated or typed text, audio, clipboard, tokens, and logs. FreeVoice does not give Gemini source code or internal architecture. Declining AI Support does not prevent you from creating a human-support ticket.

AI Support does not search the web. Earlier versions could send a separate, general Android or Google Play query for Google Search grounding. Those queries excluded raw questions, identity, account data, diagnostics, and ticket text. Google's grounding terms provide a 30-day retention period for those requests.

Human-support tickets. Before a ticket is submitted, the app shows a preview that you can edit. You choose which AI messages to include and can remove all optional information. A single switch controls inclusion of your account email and user identifier; diagnostics are visible and removable; and you may select up to three re-encoded JPEG/PNG images. Tickets never automatically include audio, dictation transcripts, clipboard contents, payment details, tokens, arbitrary files, or logs. Approved ticket text may be sent to Gemini once to produce an English summary for support. Personal ticket content is stored in Firebase and in deletable comments in a dedicated private GitHub repository accessible only to the support operator. The GitHub Issue shell contains only an opaque ticket identifier, labels, and a machine marker.

Previous support conversation. The earlier Telegram-backed support channel is read-only in updated apps. While compatibility remains for older app versions, messages sent through that channel, together with the account user identifier and email, are forwarded to the operator through the Telegram Bot API. Existing history is not copied into a new ticket unless you explicitly select content. Support data is used only to answer support requests, not for advertising or automated model training.

Android beta access email. If you email us to request Android beta access, we use your email address only to administer the test and send the FreeVoice launch updates you requested. You can ask us to remove you from the tester list and delete that correspondence at any time.

Website requests. If you visit the FreeVoice website, Firebase Hosting processes normal HTTPS request data such as IP address, browser user agent, requested path, timestamp, and security diagnostics to deliver and protect the site. The website does not use advertising trackers or remotely hosted fonts.

Who processes your data

The following service providers receive data for the purposes described in this policy. Their own terms also cover processing they perform independently, such as payment records, service security, and abuse prevention:

We do not sell your personal data, and we do not use it for advertising.

Data retention

Your choices and rights

Depending on where you live (e.g. the EEA/UK under GDPR, or California under CCPA/CPRA), you may have rights to access, correct, or delete your data, and to object to certain processing. You may also have rights to restrict processing, receive a portable copy of data, withdraw consent without affecting earlier lawful processing, and complain to your local data-protection authority. Contact us to exercise these rights; we may ask for proportionate proof that the account is yours. We do not sell or share personal data for cross-context behavioral advertising.

Legal bases for EEA and UK users

We process account, dictation, personal Dictionary, allowance, and purchase status as needed to provide the service you request. We process attestation, minimal diagnostics and service-security logs for our legitimate interests in preventing abuse and operating a reliable product. We process AI Support and human/legacy support messages to answer your request, and retain records where a legal, tax, accounting, or fraud-prevention obligation requires it. Where local law requires consent for a particular optional technology, it must be obtained before that technology is used. The current Android app does not offer an in-app Analytics consent or opt-out setting. You may object to processing based on legitimate interests by contacting us.

Children

FreeVoice is not directed to children under 13 (or the equivalent age in your jurisdiction) and we do not knowingly collect their data.

Security

Data in transit is encrypted (TLS). Authentication, your synced Dictionary, AI Support, tickets, and ticket images are protected by Firebase access controls scoped to your signed-in account. The support tracker is a dedicated private GitHub repository accessed through a one-repository GitHub App.

International transfers

Our processors operate globally; your data may be processed in countries other than your own, including the United States and the European Union. For EEA/UK transfers, we rely on the transfer mechanisms made available in the applicable processor terms, such as adequacy decisions or standard contractual clauses. Contact us if you want information about the mechanism relevant to your data.

Changes to this policy

If we change how we handle data, we will update this page and revise the effective date above.

Contact

rota2000.2@gmail.com