The biggest problem we have is that the source files are full of typos and errors. The Bible was missing verses, for example, so I had to copy-paste them in. The books are so large that the AI sort of wheezes trying to find mistakes, and I have to really scan over the whole thing.

We might just ram them all through the converter and then edit the events that are wrong, afterward, tho, since we can replace them in-place. Would get us further, faster.