Text to Speech – Convert Text to Natural MP3 Audio

Turn typed, pasted, or imported text into natural downloadable MP3 speech in English, Spanish, French, Chinese, Japanese, or Korean.

Type or paste a script, pick one of six languages, and this tool sends your text to a cloud speech-generation service to produce a downloadable MP3 — unlike most of the tools on this site, the conversion itself does not happen entirely in your browser. The result always comes back as an actual MP3 file, whatever format the speech service itself returns internally.

How to convert text to speech

  1. Type or paste your script into the editor, drag a .txt or .md file onto it, or click Import text (files up to 100 KB).
  2. Choose English, Spanish, French, Chinese, Japanese or Korean from the Language menu; the voice for that language is selected automatically.
  3. Click Generate MP3; the text is sent to Cloudflare Workers AI and the finished audio comes back to the page.
  4. Listen in the built-in player, then click Download MP3 to save the file.

This one leaves your device — here's exactly how

The text you enter is submitted over HTTPS to a server endpoint backed by Cloudflare Workers AI, which is what actually synthesizes the audio; there is no offline or fully local mode for this particular tool. The interface states this plainly before you generate anything, and the stated policy is that neither the submitted text nor the generated audio is retained by the service afterward — but the submission itself is real and unavoidable, since speech synthesis of this quality isn't something a browser tab can do entirely on its own hardware.

Language and voice selection

Six languages are available — English, Spanish, French, Chinese, Japanese and Korean — and the voice used for each is chosen automatically by the service for clear pronunciation in that language; there's no manual voice picker or list of named voices to browse. If you need a specific voice identity or a language outside these six, this tool's scope stops at these options.

Getting text in

You can type directly into the editor, or import a plain-text file with a .txt, .md, or .markdown extension up to 100KB in size. Input is capped at 5,000 characters total; a live counter tracks both word and character count and estimates roughly how many seconds of speech the current text will produce, at a rough rate of two and a half words per second. Importing a file longer than the character limit truncates it to the first 5,000 characters and flags that truncation happened rather than silently dropping the rest.

Why the download is always an MP3

After the service returns audio, the tool checks the actual bytes it received — not the filename or declared content type — to see whether they're already MP3. If they're not, the result is decoded locally in the browser and re-encoded to MP3 through the same shared encoder this site's other audio tools use, so what lands in your downloads folder is reliably a playable MP3 regardless of what container the API happened to send back.

Free speech synthesis runs 20 times a day at up to 5,000 characters per go; Pro drops the daily count.

What text to speech is useful for

Converting scripts longer than 5,000 characters

Each conversion accepts up to 5,000 characters, so split a longer script at paragraph or sentence breaks, generate one MP3 per part, and drop the parts into the Audio Joiner in order to get a single continuous file. Every part uses one of the free tier's 20 daily conversions, so a 20,000-character script takes four.

Questions people ask about Text to Speech Converter

Is my text kept private, or does someone else see it?
Your text is transmitted to Cloudflare Workers AI over an encrypted connection to generate the audio — it does leave your device, unlike most tools here. The stated policy is that the text and resulting audio are not saved by the service afterward, but if your content is genuinely sensitive, treat any cloud-based text-to-speech tool, including this one, as something the text does pass through, not something processed entirely on your machine.
Can I choose a specific voice, like a male or female narrator?
No — voice selection isn't exposed in this tool. Choosing a language automatically selects an appropriate voice for that language on the service side, and there's currently no way to pick among multiple voice options within a language.
What happens if I paste in more than 5,000 characters?
The textarea itself enforces the 5,000-character limit as you type, so you can't exceed it directly. Importing a text file that's longer than the limit truncates it to the first 5,000 characters and shows a message telling you that truncation happened, rather than silently cutting off the rest of your script.
Why did generation fail with a 'too many requests' message?
The speech service applies rate limiting, and this message appears specifically when you've hit that limit — it's a temporary throttle, not a broken request. Wait a few minutes before trying again; retrying immediately in a loop will just keep hitting the same limit.
Is this text to speech free?
Yes. The free tier gives you 20 conversions per 24 hours, each up to 5,000 characters, with no account or sign-up. A conversion is only counted when audio actually comes back, and Pro removes the daily count.
Can I use the MP3 on YouTube or for commercial projects?
BrakeTools doesn't attach a licence of its own to the audio. The speech is produced by the MeloTTS model (@cf/myshell-ai/melotts) running on Cloudflare Workers AI, so check that model's licence and Cloudflare's terms before commercial use, and make sure you hold the rights to the text itself.
Can I change the speaking speed or pitch?
No. Language is the only setting; there are no speed, pitch or voice controls. If you need a faster or slower version, the Nightcore Player can change the MP3's playback speed, but the pitch rises or falls together with the speed.
How do I download the audio as MP3?
Once Generate MP3 finishes, a player and a Download MP3 button appear. The file is named braketools-speech with the language code and a timestamp, and it is always MP3 because any other format the service returns is re-encoded in your browser first.

Related tools on BrakeTools

Content reviewed 2026-09-27.