The model I think is fairly self explanatory.

Special Note Due to the limited Dataset there will likely be errors when generated with a strong strength. I didn't play around with it too much but it works fine at a strength of .6, and likely some mild increases as well to that.