Privacy Policy

Last updated: 6 June 2026

1. Controller

The controller responsible for data processing in connection with this app is:
Felix Krebs
Rotweg 38, 71686 Remseck am Neckar
Email: xdcobra11@gmail.com

2. Overview

Voice Lab: STT, TTS & More (listed on the App Store and Google Play; in-app branding: VoiceLab) is an on-device speech toolkit for exploring features such as speech-to-text, text-to-speech, translation, voice cloning, voice activity detection, diarization, enhancement, and separation. You can use the app without registration and without creating a login account. This privacy policy explains which data is processed when you use the app and for which purposes.

The app does not provide login accounts or user profiles. We do not ask for your name or email to use the app. The app creates a persistent random device identifier stored in the iOS Keychain (or the Android equivalent secure storage). This identifier is not your name or email. It is used to link subscription restore (RevenueCat), anonymous usage metering, crash diagnostics, and optional feedback on the same device over time.

Third-party advertising SDKs may additionally process identifiers such as the Advertising ID (IDFA on iOS) when you consent, as described in section 3.2.

3. Data collection and processing

3.1 Data processed locally by the app

The app processes the following data locally on your device:

• Audio files or recordings you provide: Audio is processed on-device for speech tasks such as STT, TTS, VAD, diarization, enhancement, and separation.
• Model files and metadata: Downloaded model files, checksums, and local manifest data (e.g. download timestamp, last-used timestamp) are stored locally to manage offline use.
• Model settings and app preferences: Selected models, decoding parameters, and UI preferences are stored locally.

These locally stored data are not accessed by us as the provider and are not transmitted to our own servers. We only process personal data if you contact us directly, for example via email.

We do not train machine learning or speech models on your audio, transcripts, synthesized speech, or other session content. On-device inference uses pre-trained models you download; your recordings and generated outputs are not sent to us or to third-party model hosts for training or model-improvement purposes.

3.2 Advertising via Google AdMob

This app uses Google AdMob to display ads. For users in the EEA/UK/CH, the service is provided by Google Ireland Limited; in other regions, it may be provided by Google LLC, 1600 Amphitheatre Parkway, Mountain View, CA 94043, USA.

Data that may be processed by Google AdMob in particular include:

• Advertising ID of your device (e.g. Android Advertising ID)
• IP address
• Device information (device type, operating system version, language settings)
• App usage data (e.g. which ads are shown or clicked)
• Approximate location data (e.g. derived from the IP address), where applicable

Purposes of processing by AdMob: Serving and measuring ads, reach measurement, fraud prevention, improving the services and, where applicable, serving personalised ads based on your usage and interests.

Data sharing: The data listed above are transmitted to Google (AdMob) as a third party for advertising, measurement, and fraud-prevention purposes.

Legal basis: Where required (in particular for users in the European Economic Area, the United Kingdom and Switzerland), we obtain your consent before using personalised advertising. In those cases, processing is based on your consent (Art. 6(1)(a) GDPR). In other cases, processing is based on our legitimate interest in financing and operating the app through advertising (Art. 6(1)(f) GDPR).

For users in the EEA/UK/CH, consents are obtained in accordance with Google’s EU User Consent Policy via an appropriate consent framework (e.g. Google User Messaging Platform / UMP SDK or a compatible Consent Management Platform). If you do not give consent to personalised ads, only non-personalised ads are shown or ads are suppressed.

iOS tracking permission (ATT): On iOS, access to the advertising identifier (IDFA) is additionally subject to Apple’s App Tracking Transparency (ATT) permission prompt. If you do not grant permission, tracking across apps/websites for advertising purposes is restricted by iOS.

Changing your choice: You can change ad privacy choices in the app at any time via Settings → Advertising → Privacy Options, and on iOS also via system tracking settings.

Google privacy information:
Further information on how Google processes data in connection with AdMob can be found at:
https://policies.google.com/privacy
Information on how Google uses data from apps and websites that use Google services can be found at:
https://policies.google.com/technologies/partner-sites

Opt-out from personalised ads:
You can disable personalised advertising, for example in the Android settings, such as:
Settings → Google → Ads → “Opt out of personalised advertising” (labels may differ depending on the Android version). You can also reset your advertising ID in your device settings.

3.3 Crash and error reporting (Firebase Crashlytics)

This app uses Firebase Crashlytics, a service operated by Google, to collect crash reports and non-fatal error reports. This helps us improve the stability of the app and fix bugs, especially in combination with the various speech models and configurations the app supports.

Data that may be processed by Crashlytics includes:

• Crash and error reports (stack traces, error messages)
• Device and app information (device model, operating system version, app version, build type)
• Technical context we attach to reports (e.g. feature in use such as speech-to-text or text-to-speech, model type, truncated file or model paths). We do not attach names, email addresses or other directly identifying information.

Purposes: Analysing crashes and errors, improving app stability and fixing defects. The technical context (e.g. model type) helps us identify which configurations or models are affected.

Data sharing: The data listed above are transmitted to Google (Firebase) as the provider of Crashlytics. Processing is carried out by Google in accordance with its privacy policy.

Legal basis: We rely on our legitimate interest in ensuring the technical stability and quality of the app (Art. 6(1)(f) GDPR). You may object to this processing by contacting us; we will then review whether we can restrict or discontinue sending crash and error data in your case.

Firebase / Google privacy information:
https://firebase.google.com/support/privacy
https://policies.google.com/privacy

3.4 Model downloads from GitHub Releases

The app lets you download speech models directly from the official sherpa-onnx GitHub Releases page. These downloads are initiated by you and stored locally on your device so the models can be used offline.

Data that may be processed when downloading models:

• IP address
• Device and network information (e.g. device type, operating system, connection type)
• Download metadata (e.g. file requested, timestamp, size)

To verify integrity, the app may compute cryptographic checksums (SHA-256) of downloaded model files on your device. These checksum calculations are performed locally and are not transmitted to our servers.

Downloads are served by GitHub (GitHub, Inc., 88 Colin P Kelly Jr St, San Francisco, CA 94107, USA) or its infrastructure providers. GitHub may process the data above in accordance with its privacy policy.

GitHub privacy information:
https://docs.github.com/en/site-policy/privacy-policies/github-privacy-statement

3.5 Usage metering and configuration (Firebase Firestore)

To enforce free-tier usage limits and keep limits consistent across reinstalls on the same device, the app syncs anonymous usage counters with Google Firebase Firestore. These counters reflect how much of the free monthly budget you have used (for example speech minutes and text characters). They do not include audio recordings, transcripts, synthesized speech, or other content of your sessions.

Data processed may include:

• The persistent device identifier (see section 2) as the document key
• Numeric counters per calendar month (speech minutes used, text characters used, including free-tier and total tallies)
• Timestamps of the last sync
• Technical write identifiers used to avoid duplicate counting when offline

The app may also read remote configuration from Firestore (for example app_config/limits) to adjust free-tier limits without requiring an app update. This read does not transmit your audio or text content.

Purposes: Operating the free tier fairly, preventing abuse, and applying published usage limits.

Legal basis: Legitimate interest in operating the free tier and performing the contract for free-tier use (Art. 6(1)(b) and Art. 6(1)(f) GDPR), where applicable.

To request erasure of Firestore documents linked to your device identifier, contact us using the email address in section 12.

3.6 In-app feedback (Firebase Firestore)

If you choose to send feedback (for example from Settings or via thumbs up/down on a result screen), the app transmits only what you type together with technical metadata. Feedback is private support input and is not published in the app or shown to other users.

Data processed may include:

• Your comment text (subject to client-side profanity and spam checks)
• The persistent device identifier (section 2)
• Feature and model identifiers, app version, platform, and feedback type
• Timestamps

No audio files or transcripts are attached to feedback submissions. Storage location: Firestore feedback collection.

Purposes: Improving the app, investigating issues, and understanding feature usage.

Legal basis: Your consent when you submit feedback (Art. 6(1)(a) GDPR) and legitimate interest in product improvement (Art. 6(1)(f) GDPR).

3.7 Hugging Face and custom model sources

You may initiate downloads of speech models from Hugging Face (huggingface.co) or from user-configured HTTPS URLs (custom model catalogs). These downloads are started by you; model files are stored locally on your device.

Data that may be processed when you download from these sources includes:

• IP address and request metadata (handled by the remote host and its infrastructure)
• Download URLs and file identifiers you request
• An optional Hugging Face access token you enter in the app, stored locally on your device for authenticated downloads. This token is not an app login account or user profile.

No audio from your sessions is uploaded to Hugging Face or custom catalog hosts through these download flows. You are responsible for complying with each model’s license and the host’s terms of use.

Hugging Face privacy information:
https://huggingface.co/privacy

4. Permissions

The app may request the following permissions depending on the features you use:

• Internet access: To fetch resources and serve ads.
• Network status: To check whether a network connection is available and which type of connection is used.
• Microphone (optional): Only if you choose to record audio—for example live speech-to-text, voice cloning reference clips, or other on-device speech features. Audio is processed on your device.
• Local network (optional): If you configure a model catalog or download URL on your local network (for example a home or office server), the app may connect to that address. This is initiated by you; we do not scan your network.

Permissions can be reviewed and revoked in your device settings at any time.

5. Disclosure of data to third parties

We only disclose your personal data in the following cases:

• Google AdMob: As described above, certain data are transmitted to Google in order to serve and measure advertising.
• Google Firebase (Crashlytics): Crash and error reports together with technical and device information are transmitted to Google as described in section 3.3 (Crash and error reporting).
• GitHub (model downloads): When you download models from the official sherpa-onnx GitHub Releases page, your device connects to GitHub servers and transmits data necessary to deliver the files (see section 3.4).
• Google Firebase (Firestore): Anonymous usage counters, remote limit configuration reads, and optional feedback are transmitted to Google as described in sections 3.5 and 3.6.
• Hugging Face and custom HTTPS catalogs: When you download models from sources you configure, your device connects to those hosts as described in section 3.7.
• Legal obligations: Where we are legally obliged to do so, we may disclose data to authorities or other third parties, for example in the context of statutory disclosure obligations.

6. In-app purchases and subscriptions (RevenueCat / StoreKit / Google Play Billing)

We offer optional paid plans (for example monthly subscription or lifetime purchase). Payments are processed by Apple App Store and Google Play. We do not receive or store your full payment card details.

• Data processed for purchase handling: Store transaction status, product identifiers, entitlement status, and technical receipt-related metadata.
• Device identifier: RevenueCat receives the app’s persistent device identifier as appUserID to correlate entitlements and support restore purchases across reinstalls on the same device. This is not a login account.
• Processors/providers: Apple, Google, and RevenueCat (as subscription infrastructure provider) may process technical purchase metadata to validate purchases and restore entitlements.
• Purpose: Provide paid features, verify purchases, prevent fraud, and support restore purchases.
• Legal basis: Art. 6(1)(b) GDPR (contract performance) and, where required, Art. 6(1)(f) GDPR (legitimate interest in secure billing operations).

RevenueCat privacy information: https://www.revenuecat.com/privacy

7. Storage periods

• Local app data: Local configurations, saved audio and settings are stored only on your device until you delete them within the app or uninstall the app.
• Data processed by AdMob / Google: The storage periods for data processed by Google are determined by Google’s policies. You can find details in Google’s privacy policy (see link above).
• Crash and error reports (Crashlytics): Crashlytics data are stored and processed by Google in accordance with Firebase's data retention and privacy practices. We use the reports for analysis and bug fixing; we do not retain a separate copy of raw crash data beyond what is available in the Firebase console.
• Usage metering and feedback (Firestore): Monthly usage counters and optional feedback are retained in Firebase for as long as needed to operate limits and support. You may request erasure of device-linked Firestore data by contacting us (section 12).
• Communication via email: If you contact us by email, we store your request and the contact details you provide for as long as necessary to handle your request and as required by statutory retention obligations.

8. Your rights

You have the following rights under the GDPR with regard to your personal data:

• Right of access (Art. 15 GDPR)
• Right to rectification (Art. 16 GDPR)
• Right to erasure (Art. 17 GDPR)
• Right to restriction of processing (Art. 18 GDPR)
• Right to object to certain processing (Art. 21 GDPR)
• Right to data portability (Art. 20 GDPR)
• Right to lodge a complaint with a supervisory authority (Art. 77 GDPR)

To exercise your rights with respect to processing carried out by us, you can contact us at any time using the email address above. You may also exercise your rights regarding data processed by Google (AdMob, Crashlytics) directly vis-à-vis Google.

9. Data security

We take appropriate technical and organisational measures to protect your data against loss, misuse and unauthorised access. This includes, in particular, limiting our own data processing to what is necessary and restricting access to any administrative interfaces.

Data transmitted to third parties such as Google AdMob and Firebase Crashlytics is sent using encrypted connections (TLS/HTTPS).

10. Children

This app is not directed at children under the age of 13 and is designed as a technical speech toolkit. We do not knowingly collect personal data from children under 13. If you are a parent or guardian and become aware that your child is using the app and has provided personal data, you can contact us so that we can review the situation.

11. Changes to this privacy policy

We may update this privacy policy from time to time to reflect changes in law, technical developments or changes to the app. The current version of this privacy policy is always available at the published URL. In the event of material changes, we will provide an appropriate notice within the app.

The “Last updated” date at the top of this document will be adjusted whenever we make material changes to this privacy policy.