Difference in output with 2 different A.I. scripts

As I've written in my previous post here:

https://steemit.com/music/@the.journal/new-project-artificial-intelligence-music-composer

I started this new A.I. project, and I got time to test 2 different python scripts that use the same tensorflow A.I. neural network logic but with different results.

The train dataset was the same but the outputs quite different in my opinion.

THe first example come from a more basic script, faster in training the model but less complex in the output.

The second one uses MAGENTO state of the art developed by google script with some minor custom changes made by me to control the output.

As I said it took far away longer to train the same dataset but the output sounds better for me.