Why is macOS text-to-speech still so horrible in 2025?

It's 2025 and we have flawless AI voice generators, yet MacOS and Safari still uses some eerie, robotic version from 2015. My question is - WHY?