Inknarrate and desktop text-to-speech apps
There are already plenty of programs that run open voice models on your computer, and many of them are good. Inknarrate does not claim to replace them. It is for the people who would rather not install anything. This page says where it is better, where it is not, and how to choose.
What a web page does better
- It is just a page
- Open it, use it. Nothing to download, install, update or uninstall, and it behaves the same on Windows, macOS, Linux and Chromebooks with a modern browser.
- It stays small
- The app itself is about 16 MB, downloaded once. A voice model adds roughly 28 MB (Kitten TTS), 92 MB (Kokoro) or 15 to 110 MB per voice (Piper). Many desktop programs bundle their own runtimes and models and take several gigabytes, and some feel heavy to start and to use.
- Fast once it is set up
- After the one-time setup, it opens at once, remembers your voices, and works offline. Sentences you already generated are kept, so listening again or exporting does not redo the work.
- Good voices on modest computers
- The models were picked because they sound good and run on weaker hardware: Kitten TTS runs even on a low-end laptop, Kokoro gives very natural English, and Piper covers Spanish and Arabic. You can run several sentences in parallel if your computer has spare cores.
- Open and private
- The code is open source under the GPL, and your text is never uploaded. See the privacy policy.
Where it is not as good
- A small choice of voices
- Inknarrate supports three model families and recommends a handful of voices. Desktop programs often support many more models and languages. (Advanced users can add any Piper voice from the public Piper collection, and the browser voice covers whatever languages your device has.)
- The setup is not one click
- You download the model files from their authors and drop them into the app. It is a few minutes of work and can feel technical the first time. The built-in tutorial, the quick start in the Models window and the guide are there to make it painless, and you only do it once.
- It lives in a browser
- Models are stored in your browser. If you clear the site's data, you add them again (the Models window shows what is stored and lets you delete single files). Very long books can use a lot of memory in a browser tab; export in parts if that happens.
Which should you pick?
- Pick Inknarrate if you want something quick and light, you do not want to install software, you care about privacy, or you are on a computer you cannot install things on.
- Pick a desktop program if you need a very wide choice of voices and languages, want to work with huge texts all day, or like a full desktop interface.
- Use both. They are not exclusive. The voice models are open, and the same Piper voices work in many programs.
Other text-to-speech programs and the voice models mentioned here belong to their respective authors. Inknarrate is an independent open-source project and is not affiliated with them.
Open Inknarrate Read the guide Offline text to speech About the project