When you have auto-regenerate enabled for your project, we will automatically check the output for any mispronunciations or unwanted audio artefacts. If we detect any, we will automatically regenerate the audio up to twice, at no extra cost.
This feature will increase the processing time, and needs to be enabled.
You can either enable it in as an option in the Convert dialog window, if you're converting your Project or Chapter in one step, or in Project settings, which will impact individual paragraph generation.
How to enable in Convert dialog:
How to enable in Project settings: