How to Process a CPU Based Kokoro-FastAPI?

[dreamerseen]

Hi, firstly I would like to say thanks for this beautiful tools to any of you who contributed in it!

Currently, in my local I'm using the latest Kokoro-FastAPI GPU, it's working very well.

Then I tried to host it on a VPS, turns out the VPS doesn't support GPU, kept on failing when installing it. Then I read that we can use the CPU version. But I found that the CPU version does return a plain JSON file, compared to the GPU version which returns the audio/mp3 directly.

The question is how can I convert the CPU version into a working audio/mp3 file?

Appreciate your support.