Sometimes it works perfectly, while 20/30% of the times the parsed article contain only images