Privacy Policy

1. Information We Collect

ReadAloud AI collects information you provide directly to us, such as when you create an account, use our services, or contact us for support. This may include:

• Email address for account creation
• Content you import (articles, PDFs, text) for processing
• Voice recordings for voice cloning (if you use this feature)
• Usage data and preferences

2. Voice Cloning Data

Voice cloning is available only to signed-in, paying accounts with a verified email address. To create a voice clone you give us two recordings: a reference recording of the voice to clone, and a short consent recording in which you read aloud a phrase that we generate for you. We compare the consent recording with the reference recording to check that they are the same speaker, and we transcribe it to check that the phrase was read. A voice recording is personal data and may be biometric data under some laws.

• Reference recording: processed by Chatterbox (an open-source voice model) running on Modal's cloud GPU infrastructure under our account, and kept on that service's encrypted storage only while the voice exists. It is not stored in Supabase Storage.
• Consent recording: stored in a private Supabase Storage bucket together with its transcript, the date, and the similarity score, for as long as the voice exists and for 12 months after you delete it, so that we can answer abuse and legal claims.
• Deleting a voice (in the app or by emailing support) removes the reference recording, the voice model data derived from it, and any cached copies on the cloning service. Audio generated with your voice and the consent record are handled as described in Section 5.
• We do not use your recordings, or audio generated from them, to train any model.
• Every file generated with a cloned voice carries an inaudible watermark, and we keep a hash (a fingerprint, not the audio) of each generated file for 90 days so that misuse can be traced.

3. Third-Party AI Services

ReadAloud AI uses the following third-party and cloud-based AI services to provide its features. The app asks for your explicit consent before sending any data to these services.

Text-to-Speech

When you use cloud-based voices, your article text is sent to the following services for audio generation:

• Kokoro TTS (self-hosted on RunPod GPU cloud infrastructure) — processes text for standard quality voices
• ElevenLabs (elevenlabs.io) — processes text for premium quality voices, subject to ElevenLabs' privacy policy

AI Summarization

When you request a content summary, your article text is sent to:

• OpenAI (openai.com, model: GPT-4o-mini) — processes text to generate summaries, subject to OpenAI's API data usage policy. Data sent via the API is not used for model training.

Voice Cloning

When you create a voice clone, your voice recording and synthesis text are sent to:

• Chatterbox (open-source model, run by us on Modal, modal.com) — processes your reference recording and consent recording and your text to create and speak with your voice clone
• Speech-to-text service (our own transcription service) — transcribes the consent recording to check the phrase

Backend & Authentication

• Supabase (supabase.com) — provides authentication, database, and file storage services

On-Device Processing

Apple's built-in voices (AVSpeechSynthesizer) process text entirely on your device. No data is sent externally when using on-device voices. You can switch to on-device voices at any time in Settings.

4. How We Use Your Information

We use the information we collect to:

• Provide, maintain, and improve our services
• Process your content into audio using the AI services described in Section 3
• Create and maintain your voice clones
• Send you technical notices and support messages
• Respond to your comments and questions

5. Data Retention

• TTS audio: Cached on our servers for up to 30 days for performance, then automatically deleted
• Voice clones and reference recordings: Stored until you delete them via the app
• Consent records (recording, transcript, score): Kept while the voice exists and for 12 months after it is deleted
• Fingerprints of generated audio: Kept for 90 days
• AI summaries: Not stored on our servers after delivery to your device
• Account data: Retained while your account is active; deleted upon account deletion

6. Third-Party Data Protection

All third-party services we use provide data protection measures consistent with industry standards. Self-hosted services (Kokoro TTS, and Chatterbox for voice cloning, which runs on Modal) run on cloud infrastructure under our accounts with encryption in transit (TLS). Third-party APIs (ElevenLabs, OpenAI) are accessed under API agreements that prohibit use of your data for model training.

7. Data Security

We implement appropriate security measures to protect your personal information. All data is encrypted using industry-standard protocols. However, no method of transmission over the Internet is 100% secure.

8. Your Rights

You have the right to:

• Access your personal data
• Delete your account and associated data
• Delete your voice clones at any time
• Revoke consent for AI data sharing (limits you to on-device voices)
• Export your data
• Opt out of marketing communications

9. Website Visits and the Homepage Demo

When you visit readaloudai.org we count the page load, without cookies and without any script running in your browser. For each visit we record:

• The page you loaded and the time
• The website that sent you here (the site name only, not the address of the page you came from)
• Your browser's user-agent string
• Your approximate location (country and city), worked out from your IP address
• An anonymous visitor ID made from your IP address, browser and the date, which changes every day, so we cannot follow you from one day to the next

We do not store IP addresses ourselves. They are sent to our analytics provider, PostHog (posthog.com), only so it can work out the approximate location. We do not count requests that look like bots or automated checks.

When you play the voice demo on the homepage, we also record which example sentence you chose (or that you typed your own text and how long it was), how long the first audio took, how much audio played, and whether it finished or failed. The text you type is sent to our voice server to make the audio, but it is not included in these analytics records.

10. Contact Us

If you have any questions about this Privacy Policy, please contact us at privacy@readaloudai.org