[antirez]

Yep, makes sense... Something like 20 years ago I experimented with encoder/decoder models for lossy images compression and it worked very well, but it's a completely different domain indeed, where there aren't single local concentration of entropy that messes with the whole result.

[mikaraento]

I guess text in images would be similar, and is indeed where image generation models struggle to get the details right.

E.g., making a greeting card with somebody's name spelled correctly.