Absolutely! With visual content being predicted for popularity, a model can be made which produces quality content based on the training data of existing popular content from different sources.