I built a text to speech app that will take any text and read it. Once read, the results are cached so you never have to re-engage the model again. This is all fine. But when I ask codex to genera...