i built this on top of my own stack but the code is MIT for everything related. kokoro was pretty fun to explore, i'll likely build something similar for other models. if you have a particular preference, let me know and i'll take a look at it. the specific kokoro code i wrote to enable this is here: https://github.com/wlejon/brosoundml the models, including the bridge model i trained are here: https://huggingface.co/datasets/wlejon/brosoundml-data if you like it enough to want to try...