/jp/ - Otaku Culture

>>27647774
For the record, these kinds of models indeed work better the more metadata there is, so having information like speaker on each line is helpful. The optimal input would be a format roughly like this:

The background is of a green field with a blue sky.
Keiichi (smiling): Hi Rena!
Rena (confused, blushing): Hauu~

Here the text inside parens should be a human readable representation of a sprite like Re_se_na_a1.

Here's a new ciconiafied version of Shannon so you know which anon I am.