← ReadTune home

Plain-language privacy

Your reading profile
stays yours.

ReadTune has no account, and the extension has no analytics or telemetry. Its core reading tools work locally. This page explains what it stores and when an optional feature can send text away from your device. The public website is covered separately at the bottom.

What stays on your device

ReadTune stores your reading profile, calibration history, per-page reading position and highlights, the dyslexia-friendly-menus preference. These are saved in Chrome's local extension storage, on your device. There is no ReadTune account.

Reader View and "Restyle this page" process the page locally to reformat it. PDF mode extracts text locally in your browser, and the PDF file itself is never uploaded. If you later choose a cloud-routed AI action inside PDF mode, only the extracted text needed for that AI action can leave the device, as described below.

Read-aloud

Read-aloud uses Piper by default, a neural voice that runs entirely on your device. The default voice ships inside the extension, so it works offline with nothing to download. If you pick one of the other on-device voices in the Reading Lab, its model file is downloaded once from Hugging Face and cached on your device; the text you have read aloud is never uploaded for those on-device voices.

If you choose the optional Premium voice, each sentence as it is read to you is sent to ReadTune's own text-to-speech relay, which forwards it to a third-party voice provider and returns the audio. Only the sentence being spoken — and, prepared a moment ahead, the one after it — is sent; not the whole page. The relay does not store the text or generated audio. If it is unavailable, read-aloud falls back to the on-device voice automatically.

If instead you choose the optional ElevenLabs voice, the passage you ask to hear and your own ElevenLabs API key are sent directly from your browser to ElevenLabs to generate audio. Your key is stored only in Chrome's local extension storage on your device. Choosing Remove key also removes ReadTune's optional permission to contact the ElevenLabs API.

Talk to type (dictation)

If you use the dictation feature, ReadTune turns on Chrome's built-in speech recognition. Chrome sends microphone audio to Google's speech service to transcribe it — this is the browser's own engine, not ReadTune's. ReadTune does not record, store, or transmit the audio or transcript itself. Dictation is disabled for password fields.

AI reading help

Reader View and PDF mode have an optional Ask AI rail. Merely opening it sends nothing. When you explicitly request Summary, Ask, Define, Explain, or Annotate, ReadTune first uses on-device AI when a compatible Chrome model is already ready.

Otherwise, the relevant article text — or relevant locally extracted PDF text — is sent to ReadTune's AI relay, which forwards it to a third-party AI model and returns the result. A typed Ask question is sent with the selected context. The PDF file itself is never sent. A question you type can be more personal than reading text, so only send what you're comfortable sharing. AI output can be wrong and is labelled as approximate.

Only article summaries are eligible for ReadTune's shared 30-day response cache. Typed Ask answers, Define/Explain results, annotations, and other interactive AI responses are not written to that cache. Cached summaries are not tied to a ReadTune account. ReadTune does not send the article URL to its AI relay.

The "Simplify" rewrite pill is on-device only for now. If your browser doesn't already have a ready model, it says so rather than sending your selection to ReadTune's relay.

Service providers and infrastructure

When you explicitly use an optional network feature, the current route may involve Vercel (hosting the site and ReadTune relays), Upstash Redis (when configured, for the shared summary cache and short-lived rate-limit counters), OpenRouter and its configured model/voice providers, Ollama Cloud as an optional AI route, Deepgram or Fish Audio through OpenRouter, and Cartesia as a direct Premium voice fallback. ElevenLabs, Hugging Face, and Google/Chrome speech recognition are used only for the specific optional features described above. Provider availability and routing can change; this policy will be updated before a materially different data-sharing route is shipped.

Abuse prevention and request metadata

ReadTune's relays use the requesting network address only to derive a short pseudonymous HMAC fingerprint for per-client rate limiting. The raw address is not written into ReadTune's Redis rate-limit key, and minute-level counters expire automatically. Vercel and other infrastructure providers may separately process ordinary request, security, and network logs needed to operate their services.

Permissions

Chrome Web Store Limited Use

ReadTune uses information only to provide or improve its disclosed reading-accessibility features and related security or operational purposes. It does not sell user data, use it for advertising, creditworthiness, or lending, or transfer it for unrelated purposes. The use of information received from Chrome APIs adheres to the Chrome Web Store User Data Policy, including the Limited Use requirements.

What ReadTune does not do

The extension does not sell or share data, run analytics or telemetry, create an account, or send your browsing history to its creator. It does not use data for advertising, creditworthiness, or lending. The extension is open source and can be inspected on GitHub.

This website

The public marketing pages at readtune.tech are hosted on Vercel. ReadTune does not load a pageview analytics SDK, advertising pixel, or product-tracking script on these pages. As with ordinary web hosting, Vercel may process basic request and security logs needed to deliver the site; ReadTune does not use those logs to build reader profiles or connect website visits to extension activity.

Nothing you read inside the extension is visible to this website. Reader View, "Restyle this page", PDF mode, read-aloud, calibration, and highlights do not load website analytics.

Contact

For questions or bug reports, use the project's GitHub Issues.

Last updated: September 2026