Output length of RNN models

Hi, is it necessary to have the number of words in the output to be the same as the number of layers in the RNN architecture?

This image shows that if we use different sampling techniques, we get different results. But how can we get 2 different lengths of outputs from the same architecture?

First, please note that these are not the number of layers, they are cells that are stacked together. Or you can say one layer repeated (recurrent) several times. Moreover, many-to-many can produce different lengths of output.