^ A thought that everyone has had at one point when processing human text before learning the hard way (like end of sentence detection). :P

The difference is that even weak LLMs are good at magically doing this, so I wonder what the problem is for the TTS mentioned above.