This is a style model trained off riley_moore's work

I would put negatives such as: English text, speech bubble