Text to Speech
This free text to speech tool reads any English text in a natural AI voice. Press Play to listen while each sentence lights up, or Download to save an MP3. The voice runs on your own device, so there is no signup, no character limit, and your text is never uploaded.
How do you turn text into speech for free?
Paste the text and press Play: it reads any English text aloud in a natural AI voice, and Download MP3 saves it as a file. No signup, no daily limit and no watermark.
- Natural voices, not robotic ones. The voice is Kokoro, a neural model that reads whole sentences with natural rhythm, pauses at commas and periods, and the rising tone of a question.
- Play starts as soon as the first sentence is ready. Speech is created sentence by sentence while the previous one is already playing, and the sentence being read lights up. Click any sentence to jump there.
- Unlimited MP3. Press Download MP3 for the audio file. Anything you already heard is saved instantly. Up to 200,000 characters at a time (a short book), with no daily limit, no signup and no watermark.
- Private by design. The voice runs on your own device, so your text is never uploaded.
- 18 voices, and your own mixes. American and British, female and male. Blend two voices into a new one, and tune speed, expressiveness and the pause between sentences.
Step by step
- Type or paste the text, open a .txt file, or press Try a sample text.
- Pick a voice. Heart is the most natural; Emma is the best British voice; Michael and George are good male voices.
- Press Play to listen, and Pause or Stop at any time.
- Press Download MP3 to save the audio.
What do the voice controls do?
Each control changes the voice itself, not just the playback.
| Control | What it does |
|---|---|
| Mix | Blends two voices into a new one. Every voice is a set of 256 numbers describing its sound; the mix is a weighted average of the two |
| Expressiveness | Pushes the voice away from a neutral average for livelier intonation, or toward it for a calmer read. Livelier also sounds slightly higher and quicker |
| Speed | From half speed to twice as fast |
| Pause between sentences | From no pause to over a second, for narration or study |
At the default settings the voice is exactly the original Kokoro voice.
How does the voice run on your device?
Because the voice itself comes to your device. The first time, the browser downloads the voice model once: 310 MB on computers whose graphics card the browser can use, or a lighter 88 MB version elsewhere. The browser keeps it, so every later visit starts right away.
With a graphics card, speech is created several times faster than it plays, so listening is smooth. Without one, it is created a little slower than it plays, and there can be short pauses between sentences. The MP3 is the same quality either way.
What we tested
We pasted a 129-word memo full of the things that trip readers up: a date (Jan. 5), a time (3.50 p.m.), a price ($12,400), a percentage, a quote and abbreviations, and pressed Download MP3 with the Heart voice.
| What we checked | Result |
|---|---|
| How the text was split | 13 sentences, none broken at Dr., p.m. or Jan., and the quote kept with "asked Maria" |
| The file | a standard MP3, 56.6 seconds, 885 KB (128 kbps, mono) |
| Time to make it | 42 seconds on a laptop with a graphics card, loading the voice included |
| Speaking pace | about 141 words per minute at speed 1.0; the page estimated 57 seconds before making it |
| Name of the file | taken from the first words of the text |
Listen before you publish: a rare name or an unusual abbreviation can still come out wrong, and stopping on that sentence to rephrase it is the quickest fix.
What can't it do?
- Speak other languages. The voice was trained on English, so other languages sound wrong.
- Read PDFs or web pages directly. Copy the text in, or open a .txt file.
- Clone a voice or add emotions on command.
To check how long a script will run before recording, the word counter shows speaking time too.
Frequently asked questions
How do I convert text to speech?
Type or paste your text, choose a voice and press Play. It starts speaking as soon as the first sentence is ready, usually a second or two with a graphics card, and highlights the sentence being read. Click any sentence to start there. Press Download MP3 to save the audio as a file.
Can I download text to speech as an MP3 for free?
Yes, with no character limit, no daily cap and no signup. Anything you already listened to is saved instantly; the rest is created on the spot with a progress bar. The MP3 is made on your device, so there is no watermark and nothing waits on a server queue.
Why does this voice sound natural?
It uses Kokoro, an open neural voice model with 82 million parameters. It reads a whole sentence at a time, so the rhythm, the pauses at commas and periods and the rising tone of a question come out the way a person would say them. The model is small enough to run inside your browser.
Why does the first Play take a moment?
The first time, your browser downloads the voice itself: 310 MB on computers whose graphics card the browser can use, or a lighter 88 MB version elsewhere. The browser keeps it, so later visits start right away. It is a one-time download, like installing an app, and never happens again on that browser.
Is my text private?
Yes. The voice runs inside your browser tab on your own device, so the text is never sent anywhere. The only downloads are the voice model and the code that runs it, fetched once from public open-source servers. That makes it safe for private notes, emails and unpublished drafts.
Which voice should I pick?
Heart and Bella are the most natural American voices, Nicole is soft and calm, and Emma is the best British voice. For a male voice, try Michael or Fenrir in American English, or George in British English. You can switch voices between sentences by stopping, choosing another and pressing Play again.
Is there a character limit?
Up to 200,000 characters at a time, about a short book. Long texts are read sentence by sentence, so a whole article works like a single line. With a graphics card the voice is created faster than it speaks; on older computers there can be short pauses between sentences.
Can I use the audio in videos and presentations?
The Kokoro model that creates the voice is released under the Apache 2.0 license, which allows commercial use, and no watermark is added. Platforms have their own rules about synthetic voices, such as labels for AI generated content, so check them before publishing.
What languages does it speak?
American and British English. The AI voice was trained for English, so text in other languages is read with English pronunciation and sounds wrong. Names and uncommon words are usually read well, but listening before you download is the quickest way to catch the rare word it gets wrong.
Can I customize the voice?
Yes, with four controls that change the sound itself. Mix blends two voices into a new one, such as Heart with Michael. Expressiveness goes from calmer to livelier intonation. Speed runs from half to double, and Pause between sentences sets the breathing room for narration. Every change applies the next time you press Play.
Related tools
Rate Text to Speech & help shape it
This tool is free and still growing. Tell us what works, what you would change, and what is missing. Your feedback is what decides what we build next.
Built and tested by the ÆON Toolhub team. Last updated: October 19, 2026. How we test our tools