Text to Speech Online: Natural-Sounding Audio You Can Actually Download
Turn text into natural-sounding speech and download it as MP3 or WAV. How to pick a voice, write text that reads well aloud, and where synthetic voices fall short.
Updated September 15, 20266 min read
Text to speech turns written words into a spoken audio file: paste your text, pick a voice, generate, and download the result as an MP3 or WAV. The voice matters less than you would expect. Most of what makes synthetic speech sound natural or robotic is decided by the text you give it — the punctuation, the numbers, the abbreviations — and by knowing which jobs a synthetic voice does well and which it does not.
Two engines, and when to use each
Our text to speech tool puts two engines behind one text box, because the two things people want pull in opposite directions: something that speaks immediately, and something you can save as a file.
| Instant | Studio | |
|---|---|---|
| Voices | Whatever your operating system provides | 28 voices, American and British English |
| Starts | Immediately | After a one-off download of roughly 90 MB, cached after that |
| Languages | Usually many, depending on your device | English only |
| Download | No — it plays straight to your speakers | WAV, MP3, M4A, OGG or FLAC |
Use Instant to listen to something once, or for a language other than English. Use Studio for anything you want to keep: a voice-over, an audio version of a document, a clip for a presentation. System voices cannot be saved because the operating system plays them directly and gives the page no audio to record; that limitation is the reason the Studio engine exists.
Voices, speed and download formats
Studio has 20 American English voices (11 female, 9 male) and 8 British English ones, listed best first. The pronunciation dictionary behind them is American, so the British voices speak with a British accent but American vowels — audible on words like "bath" and "dance". If a voice sounds thin on your text, try one higher up the list.
The speaking rate runs from 0.5× to 2×. There is no pitch control; if a voice sits too high or low for your project, choose a different voice rather than trying to shift one. For downloads, WAV is the exact output at 24 kHz, best when the audio goes into an editor. MP3 is the smallest and plays everywhere. M4A suits Apple devices, OGG the web, and FLAC is lossless but smaller than WAV. For another format later, run the file through the audio converter.
Writing text that sounds good read aloud
Text written for the eye often reads badly aloud. The tool reads punctuation as pauses and intonation, not as words, so the rhythm of your sentences becomes the rhythm of the speech.
- Punctuate properly. Commas and full stops are where the voice breathes. A long sentence without them is delivered as one breathless run.
- Leave a blank line between sections. A paragraph break becomes a longer pause, which is what separates the parts of a script.
- Keep sentences short. A listener cannot glance back. Two sentences are easier to follow than one with three clauses.
- Check how numbers will be read. Figures are read the way a person would: 1,234 as a quantity, "in 1999" as a year, $19.99, £ and € amounts as money, 3rd as an ordinal and 25% as "twenty five percent". If a number means something unusual, write it out in words.
- Expand abbreviations. Acronyms such as PDF are spelled out letter by letter, which is right for PDF and wrong for anything meant to be said as a word. Write out "for example" and similar in full rather than relying on the abbreviation.
- Respell awkward names. Unfamiliar names are guessed from their spelling and are usually close. When one is wrong, a phonetic spelling in the text fixes it.
Paste or import the text
Type it in, or import a .txt or .md file.
Generate one paragraph first
Listen for mispronounced words and awkward pauses before committing to the whole document.
Fix the text, not the voice
Most problems are solved by a comma, a line break or a respelling.
Generate and download
Choose the format, then save the file.
Paste your text, pick one of 28 voices and download MP3 or WAV. It all runs on your device, not our servers.
Convert text to speechWhat people use it for
Voice-overs. Download WAV for editing, set the rate to match your slide or scene timings, and drop it into your video as the narration track. A voice-over rarely sits at the same level as the music under it; the audio normalizer brings it to a consistent loudness.
Proofreading by ear. Your eyes skip over a repeated word or a missing one because your brain supplies what you meant to write. A voice reads exactly what is there. Listening to a draft is one of the most effective ways to catch sentences that are too long, and it works best in a voice that is not your own.
Accessibility and listening on the move. Some people take in long documents more easily by ear, and a downloaded MP3 can be played anywhere. If you need to go the other way — a recording into text — speech to text does that, and our guide to transcribing audio covers it in depth.
The honest limits of synthetic voices
- It reads, it does not act. The delivery is clear and even. A joke, a dramatic line or real emotion will come out flat.
- Unusual words are guesses. Brand names, place names and jargon are predicted from spelling and are occasionally wrong. Listen before you publish.
- English only for downloads. The Studio voices are English. Other languages are available only through the Instant engine, which cannot be saved.
- Speed depends on your device. The work happens on your machine, so a book-length document on an old phone is slow. Split it, or run it on a laptop.
- No voice cloning. The 28 voices are the voices; the tool will not imitate a particular person.
Free and Pro
The free plan converts up to 5,000 characters at a time — roughly five to six minutes of speech at normal speed. Every voice, every speaking rate and every download format, MP3 included, is free. Pro removes the character limit and converts up to 25 documents in one batch, delivered as a single archive. If you are working to a character count, our guide to word and character limits explains how they are measured.
Frequently asked questions
Is there a free text to speech with MP3 download?
Yes. Use the Studio engine, generate the speech and download it as MP3 — or WAV, M4A, OGG or FLAC. Every voice and format is available on the free plan, up to 5,000 characters per conversion.
What is the most natural sounding text to speech voice?
Use the Studio engine; system voices vary from excellent to very robotic depending on your device. The Studio voices are listed best first, and the American voices match the pronunciation dictionary exactly. Well-punctuated text makes more difference than the choice of voice.
How many characters can I convert to speech for free?
5,000 characters per conversion, which is roughly five to six minutes of speech. Pro removes the limit and adds batch conversion of up to 25 documents at once.
Can text to speech read languages other than English?
Through the Instant engine, yes — it uses whatever voices your operating system provides, which usually covers many languages. Those voices cannot be downloaded, though. The downloadable Studio voices are English only.
Why does the first conversion take longer?
The Studio engine downloads its voice data the first time you use it, roughly 90 MB. It is cached afterwards, so later conversions start straight away. The Instant engine has nothing to download.