sqwartHome

Text to Speech

Natural-sounding English voiceovers from any text, made by an AI voice model on your own device.

  • Free
  • No upload
  • No watermark
  • No sign-up
  • AI runs on your device
How it works

How to use the text to speech tool

  1. 1

    Type or paste your English text, up to 20,000 characters.

  2. 2

    Choose a voice and a speed.

  3. 3

    Press Make speech and listen to the result.

  4. 4

    Download it as MP3 or WAV.

In detail

Natural voices, made on your device

The voices come from Kokoro, an open text-to-speech model with 82 million parameters (Apache-2.0), small enough for a browser yet natural enough to rival much larger models. It runs inside your browser: the text is turned into phonemes with eSpeak NG, then Kokoro speaks them as 24 kHz audio. The model is a 92 MB download the first time; your browser keeps it, and your text never leaves your device.

Open-source parts: Kokoro 82M (Apache-2.0, huggingface.co/hexgrad/Kokoro-82M), kokoro-js (Apache-2.0, github.com/hexgrad/kokoro), and eSpeak NG (GPL-3.0, github.com/espeak-ng/espeak-ng) for phonemes, whose source code is available at that address.

There are 12 voices here: American and British, women and men. Heart and Bella sound the most natural; Michael and Fenrir are good male voices; Emma and George are British. Long texts are spoken a sentence or two at a time and joined with short natural pauses, so a whole article becomes one file.

English only, and other honest limits

  • It speaks English only. Kokoro has voices for a few other languages, but not Hindi, and this tool offers only the English ones, which are by far the best.
  • Speed: on a typical laptop processor, about one to three seconds of work for every ten seconds of speech, so a 1,000-word script (about 7 minutes of audio) takes a minute or two.
  • Names, acronyms and unusual words are sometimes mispronounced. Spell them the way they sound ("Sqwart" as "squart") if needed.
  • There is no voice cloning, emotion control or SSML. Use punctuation to shape the pacing: commas give short pauses, periods longer ones.

Good uses

Voiceovers for YouTube Shorts, Reels and explainer videos; narration for slides and courses; listening to an article instead of reading it; quick drafts of a script before you record it yourself. The audio is yours to use, commercially too.

FAQ

Questions, answered.

Is it free with no character limit?

It is free with no account. Each run takes up to 20,000 characters (about 3,000 words); for longer texts, run it in parts.

Does it speak Hindi?

No. This voice model speaks English only. Hindi text will be skipped or mispronounced.

Can I use the audio in YouTube videos?

Yes. Kokoro is under the Apache-2.0 license, which allows commercial use, and there is no watermark in the audio.

Is my text sent to a server?

No. The voice model runs inside your browser, so your text stays on your device.

Why is the first run slow?

Your browser downloads the 92 MB voice model once (plus 31 MB for the MP3 encoder) and keeps them. Later runs start at once.

MP3 or WAV?

MP3 is about a tenth of the size and fine for videos and sharing. WAV is uncompressed, for editing in an audio app.

Keep going

More free audio tools