Does it work without internet?

Yes, once it is set up. The first visit loads the app, and the voice model files you add are stored in your browser. After that you can turn the internet off and keep generating speech.

Yes. There is no account, no subscription, no usage limit and no watermark. Inknarrate is open-source software under the GPL.

Is my text sent to a server?

No. The speech is produced by your own browser, so your text and audio never leave your computer. Details are in the privacy policy.

Which languages does it support?

English with Kitten TTS, Kokoro or Piper, and Spanish and Arabic with Piper voices. The built-in browser voice covers whatever languages your device has installed.

Can I save the audio as MP3?

Yes. With a voice model added you can download the recording as WAV or MP3, get subtitles (.srt or .vtt), or a zip with one file per sentence. The browser's built-in voice can only be played, not saved.

Does it work on a phone or an old laptop?

It is built for desktop browsers. The small Kitten TTS model and Kokoro's quantized model run on modest laptops (Kokoro worked on a 2017 dual-core laptop with 4 GB of RAM, a little slowly). Phones have not been tested, and large models may be too heavy for them.

Can I use the audio commercially?

That depends on the voice model, not on Inknarrate. Each voice has its own licence from its authors, so read the model page before commercial use.

Free offline text to speech in your browser

You can turn text into natural-sounding speech without installing anything, creating an account, or sending your text to anyone. Inknarrate does it inside a web page, and once it is set up it keeps working with the internet switched off.

What “offline text to speech in the browser” means

Most online text-to-speech sites send your text to a server, which turns it into audio and sends it back. That needs a connection, often has limits or a price, and means a company receives everything you type.

Inknarrate works the other way round. Small open voice models are stored in your browser, and WebAssembly runs them on your own computer. The text is converted to speech right there. No server is involved in making the audio, so there is nothing to upload, no daily limit, and no watermark on what you save.

How to start in three steps

Open the app and press Speak. It uses your browser's built-in voice straight away, so you hear something in seconds.
Add a natural voice, once. Press Models, download the files of a model (Kokoro is the one we recommend for English), and drop them into the box. They stay in your browser. The built-in tutorials walk you through it.
Write, listen, save. Paste or import your text, press Speak, and download the result as WAV or MP3, with optional subtitles.

Open Inknarrate Read the guide

What people use it for

Proofreading by ear. Mistakes you skim over on screen are easy to hear.
Listening to long texts such as articles, notes or chapters you imported from .txt, .epub or .docx files.
Language practice. Hear Spanish or Arabic text read aloud at a speed you choose.
Voiceovers for your own projects. Export a clean recording with subtitles. Check the licence of the voice you used first (see below).
Working offline or on a restricted computer where you can't install software.

Languages and voices

English is covered by three model families: Kitten TTS (tiny and fast), Kokoro (the most natural) and Piper. Spanish and Arabic work through Piper voices. For Arabic, adding vowel marks (التشكيل) to your text makes the pronunciation noticeably better, and the app can add them for you. The built-in browser voice covers any other language your device has installed.

You choose and download the models yourself, from their authors on Hugging Face. Inknarrate does not host or bundle them, which keeps the app small and each voice under its own licence.

Is it private?

Your text and audio stay on your device. The only outside contact is when you press a Download button for a model file, which goes to Hugging Face. See the privacy policy for the full list.

How it compares with desktop programs

Desktop programs built on the same open voice models exist and many are good. A web page is lighter and needs no installation, but it offers a smaller choice of voices. Here is an honest comparison of where each one fits.

Limits worth knowing

The choice of voices is small compared with large cloud services or big desktop programs.
Setup takes a few minutes the first time, because you add the model files yourself.
Models live in your browser's storage. If you clear the site's data you add them again (the app lets you manage and delete stored files).
Voices are good but not human. Expect natural reading, not acting.

Licences and commercial use

Inknarrate itself is free software under the GNU GPL v3 or later. The voice models are separate works with their own licences, set by their authors, and some voices have limits on commercial use or unclear terms. The app shows a note next to the suggested voices; read the model page before using audio in a product or a paid project.

Questions

Does it work without internet?: Yes, once it is set up. The first visit loads the app, and the voice model files you add are stored in your browser. After that you can turn the internet off and keep generating speech.
Is it really free?: Yes. There is no account, no subscription, no usage limit and no watermark. Inknarrate is open-source software under the GPL.
Is my text sent to a server?: No. The speech is produced by your own browser, so your text and audio never leave your computer. Details are in the privacy policy.
Which languages does it support?: English with Kitten TTS, Kokoro or Piper, and Spanish and Arabic with Piper voices. The built-in browser voice covers whatever languages your device has installed.
Can I save the audio as MP3?: Yes. With a voice model added you can download the recording as WAV or MP3, get subtitles (.srt or .vtt), or a zip with one file per sentence. The browser's built-in voice can only be played, not saved.
Does it work on a phone or an old laptop?: It is built for desktop browsers. The small Kitten TTS model and Kokoro's quantized model run on modest laptops (Kokoro worked on a 2017 dual-core laptop with 4 GB of RAM, a little slowly). Phones have not been tested, and large models may be too heavy for them.
Can I use the audio commercially?: That depends on the voice model, not on Inknarrate. Each voice has its own licence from its authors, so read the model page before commercial use.

Try it now About Inknarrate Privacy policy