Listen to private documents without uploading them
Most text-to-speech sites ask you to upload your file. This one does not: PDFs, Word documents and ebooks are opened and parsed by code running inside your own browser, and the file never leaves your machine.
The text you choose to have spoken does get sent, because that is where the voices live. The table below says exactly what crosses the wire and what does not, so you can judge it rather than trust it.
- Files parsed locally
- Nothing stored
- Free
Edit the line to hear anything you like.
What leaves your device, and what does not
| Thing | Where it goes |
|---|---|
| Your PDF, Word or EPUB file | Nowhere. Parsed in the browser by code running on your machine. |
| Text you send to be spoken | To a third-party neural voice service, which synthesizes it and returns audio. Not stored here afterwards. |
| Text you import but delete before generating | Nowhere. Only what you actually generate is ever sent. |
| The URL of an article you import | To this site's own server, which fetches the page and returns its HTML. It travels in the request body rather than the address, and nothing in the code writes it to a log. |
| Audio you transcribe | Nowhere. The speech model runs on your device; the recording is never transmitted. |
| Generated MP3 and captions | Held in your browser's memory until you download or leave. Never saved on a server unless you press Publish. |
| A recording you publish as a listen link | Uploaded on purpose, with its words, so the link can play them. Unlisted, expires on its own, and deleting it removes the audio. |
| Drafts and preferences | Your browser's local storage, on your machine. Clearing site data deletes them. |
The one honest caveat is in row two. A tool that promised complete locality while sending text to a cloud voice would be lying, and the distinction is the whole point of this page: the document stays, the selected text goes.
Why parsing locally is worth the effort
A document usually contains far more than the part you want read aloud - other sections, metadata, revision history, the names of everyone who touched it. Uploading the file hands all of that over to get one chapter spoken.
Parsing in the browser means the extraction happens where the file already is. You see the extracted text in the editor, you cut it down to the part you actually want, and only that part goes anywhere.
The one genuinely local tool
Transcription is the exception that needs no caveat at all. The speech recognition model is downloaded into your browser and runs on your own processor, so a recorded meeting, interview or voice note is never transmitted anywhere - not the audio, not the transcript. It works offline once the model has loaded. Try the transcriber.
Questions
- So nothing at all is sent to a server?
- That is not quite true, and it matters. Your file is never sent - it is parsed in your browser. But the text you choose to have spoken is sent to a third-party neural voice service, because that is where the speech is synthesized. If a document is sensitive enough that no third party may see any part of its text, do not send that part.
- Is my text stored after the audio comes back?
- Not by this site. Nothing you type or import is written to a database or kept on disk here. Drafts are saved in your own browser's local storage so a closed tab does not cost you work, and clearing your site data removes them.
- What about the transcriber?
- Audio transcription is fully local. The speech model is downloaded into your browser and runs on your own device, so a recording you transcribe is never transmitted at all - not even the text it produces.
- Do I need an account?
- To generate audio, yes - a quick sign-in with Google or with an email and password. Opening and reading your documents works without one. Signing in shares only your email address (and your name, with Google) with this site; it changes nothing about what stays on your machine.
- What is logged?
- Usage metadata, linked to your account while you are signed in: the kind of action, how many characters, which voice, which format, and how your visit began. Never your text, your documents, your audio, or the URLs you import. The privacy policy lists the fields exactly.
- Can I use this on confidential work material?
- That is your call and your employer's, not ours. The honest summary is: the file stays on your machine, the text you select for speech does not. Judge it against that, not against a marketing claim.