Unlimited text-to-speech using Kokoro-JS, 100% local, 100% open source

The entered text is not sent to any server, instead a 300MB AI model is downloaded once and used to turn any text into speech.

Source code is here: https://glitch.com/edit/#!/streaming-kokoro

I'm on crappy internet right now so tried downloading the audio instead, which doesn't work and always fails with an "error initializing file save". Does the download option still require the model to be downloaded first?