01 / Write
Paste your text
Up to 20,000 characters at a time, about 20 minutes of speech. Numbers are read as words.
Free tool
Type or paste a script and get it read aloud by a natural sounding voice, ready for a voice-over, a tutorial or a video. Pick from eight American and British English voices, a Hindi man or woman, or a Bangla man or woman, set the speed and save an MP3 or WAV. Open AI voice models run in this tab, so your text is never uploaded. No sign-up.
This runs in your browser. Your text is never uploaded.
How it works
01 / Write
Up to 20,000 characters at a time, about 20 minutes of speech. Numbers are read as words.
02 / Pick
The first time, the page downloads the AI voice, about 92 MB, and keeps it for next time.
03 / Save
Play it in the page, then save an MP3 or a WAV for your video editor.
Questions
No. The AI voice model is downloaded to your browser and speaks there, so your script never leaves your computer. There is no account and no monthly character limit.
For English, Kokoro, an open text to speech model with 82 million parameters, released under the Apache 2.0 licence. Words are turned into sounds with the misaki pronunciation dictionaries, also Apache 2.0. For Hindi and Bangla, VITS voices made by SYSPIN: the Hindi woman, Hindi man and Bangla woman under the CC BY 4.0 licence, and the Bangla man under the MIT licence. We converted them to run in the browser.
Yes. Both models' licences allow commercial use, and the audio is made on your own computer. The Hindi voices and the Bangla woman are under CC BY 4.0, which asks for credit, so name SYSPIN as the voice in your credits. Check the rules of the platform you post on about AI voices.
English, in American and British voices, Hindi and Bangla. For English, words that are not in the dictionary, like new brand names, are guessed from their parts or spelled out, so write unusual names the way they sound if they come out wrong. The Hindi and Bangla voices read their own script only: write English words in Hindi or Bangla letters, such as यूट्यूब or ইউটিউব. Numbers are read as Hindi or Bangla words. Each of these voices is a separate download of about 73 MB the first time.
It depends on your computer. In our test on a Mac it made about one second of speech every two seconds, so a one minute voice-over takes about two minutes. The Hindi and Bangla voices are faster: 10 seconds of Bangla speech took 13 seconds in the same test. The page shows each part as it is made.
No. It speaks only in its built-in voices.