This looks very cool. I really hope the results are not overly cherry-picked like Adobe's first version of the text-to-vector generation that only worked particularly well for the showcased art styles.