Is OpenAI’s DALL-E text to image good enough?