How to train Chinese tradtional vertical in Tesseract 5?

[fishfree]

This is the screenshot from jTessBoxEditor:

The provided example training files in this repo seems building a whole line of image & text pairs, other than character-by-character.
Then my questions are:

1. How to efficiently split a single image into mulitple vertical lines of text?
2. By which directon do we write the vertical text in text files, L2R or R2L?

[fishfree]

For Q1, I noticed the answer here.