Proofread by listening

What listening catches, and what it misses

Listening is very good at exposing missing or repeated words, sentences that are too long to follow, clumsy rhythm, accidental repetition of the same word in nearby sentences, and places where the logic jumps. If you run out of breath imagining yourself saying a sentence, the voice will make that obvious.

It cannot catch errors that sound correct. Their and there, its and it's, affect and effect sound the same or nearly the same, so a spoken reading will not reveal a mix-up. Punctuation, capitalization, double spaces and formatting are also invisible to the ear. Numbers are read in their spoken form, so a date or amount will sound fine even if it is the wrong one. Use listening alongside a spellchecker and a final visual pass, not instead of them.

A listening workflow that works

Paste your draft, or import it directly: Word documents, PDFs with real text, Markdown and plain text files can all be dropped onto the text box, and everything is read on your device. See Word to speech if you work mainly in DOCX. Generate, then keep your original document open in another window.

Follow along in your own document while the audio plays, not in the tool's text box. When something sounds wrong, pause, fix it in your document, and carry on. Stopping to fix every issue as you hear it is slower than marking them and fixing them together, so many people simply highlight the problem spots on the first pass.

The parts list shows the draft split into short pieces of a sentence or two. Click any part to replay from there, which is useful when you want to hear a tricky paragraph again. You can also edit a part directly and regenerate only that part to check that your rewrite reads better.

Choosing the voice and speed

The preset uses Nicole, a soft, slightly slower voice with a close-mic style that is comfortable for long listening. At 1.1× it keeps a good pace without rushing. For a first pass on a long document, 1.1× to 1.25× keeps you moving; for a final check on an important paragraph, drop back to 1× or 0.9× so nothing slips past.

A different voice can also help. If you have listened to the same draft several times, switching to Heart or a British voice such as Emma makes it sound fresh again, which is part of what makes listening work in the first place.

Private by design

Drafts are often sensitive: a cover letter, a grant application, an unpublished chapter, a client report. The tool runs in your browser and your text is never sent to a server, so nothing you paste leaves your device. After the first use it also works offline, which means you can proofread on a train or plane. Your draft is kept in the browser on this device between visits, so you can close the tab and come back to it later.

Long documents

There is no character limit, so a full thesis chapter or report is fine. Playback starts while later parts are still being generated, so you do not have to wait for the whole document. On slower devices and phones it can take longer, and the tool tells you how long it needs before playback starts. For more on this, see long text to speech.

Frequently asked questions