Can I generate/replicate an entirely new text using an autoencoder trained on a different set of text?

I am working on text generation and taking help of the following example:

https://in.mathworks.com/help/deeplearning/ug/generate-text-using-autoencoders.html

This code is running well and good. I have tested this code by giving different input text and then generating the same input text as output. Let us suppose, this trained autoencoder for text generation is named as "AE"

Now, my query is: Is it possible to generate an entiely new set of input text using the earlier trained autoencoder "AE"?

Feel free to comment if my query is not clear.