Hey,

Cool ideal to package this all in one, I was previously testing KokoroFastAPI + epub2audiobook which worked well, but this all-in one with a GUI is very cool.

One feature that is currently missing is the ability to mix voice models in Kokoro. KokoroTTS allows you to merge voice models to create new sounding voices. For example from my brief testing, the best sounding voice model to me was actually a model mixing two of the american voices: Bella and Alloy.

With the fast API docker container I would use the following format as the "voice name" to get this to work voice(weighting)+voice(weighting). E.g. af_bella(3)+af_alloy(1)
I know that other software uses voice1:weight,voice2:weight

Here is a web-based site that shows some of the possibilities. It's also worth noting that the default Kokoro "af" voice is actually a 50:50 weighted mix of Bella & Sarah so there is established precident that it can sound better than single voice options.

Keep up the cool project!