Zipf Distribution based number generation

I want to generate a popularity distribution for a small data set, which should follow Zipf law.

The available parameters are: Total number of viewers : 1 Million Total number of videos : 36

I want to associate total number of viewers to each video according to Zipf law. for instance, how many of viewers will watch video1, video2 .. and so on. Can anyone please tell me the formula or methodology.