Hard not to laugh, but you'd be surprised how often even other TTS companies are messing this kind of thing up. I think it has a lot to do with the data source they use for training, which evidently doesn't include a lot of currency amounts...

If you're curious what's possible with <.01% of the funding, check out https://rime.ai/. We train on data recorded in our studio and specifically include a lot of currency in our scripts for this very reason.

[disclaimer: one of the founders of Rime]