Show HN: Kokoro TTS model extrapolation and experimenting application

I recently saw Kokoro TTS speech added voice blending and after a bit of playing around I decided I wanted a more comprehensive way of selecting and blending voices. I used sqlite as a solution for selecting groups of voices to compare. I then use linear interpolation between the two groups to create voice samples between and beyond the existing grouping of models. This lets me experiment in creating new voices with traits that are exaggerated or diminished depending on what I am going for.

The code also allows for exporting to the tensor (.pt) and the kokoro-onnx (.bin) files for use in other applications.