Good point, thanks. And I was thinking it will show that the model can really synthesize very varied samples... ;)