Generally speaking I TTS works better broken down, this is for iterative control, and because the longer AI processes the worse it becomes in audio sense, so it would be best to break it down into sense < the character limit.

the las thing you'd want to do is produce and audio book and 15 minutes into your audio it starts slurring and being weird. and then you have to fix it by regenerating it. another tip is to create a naming and iteration for them so that you know which is version 1 verses 2.

Like with all TTS it's good practice to break down your work count to maximize the effectiveness of pronunciation, TTS is absolutely an iterative process, you can't expect it to get the pronunciation perfect and the way you want it with pace every time 100% THAT, wouldn't be fair. It would be best to use your first months limits to give the audio test scenarios of your target ideas, this way you can build check lists and fine-tune the nuances that could occur and be ready for a production cycle the next, this allows you to pre-plan a cycle and make the most of it whilst reducing errors in the future.