The time-varying amplitude envelope of a sound-file is simply a representation of how loud the sample gets at particular points. If you consider finding the loudest sample within some section of the sound-file, finding the loudest sample in the next section, and so on, then you have some idea of how we calculate the amplitude envelope of the sound.

Once we have calculated the amplitude envelope, however, we may then apply that data to another sound-file. For example, if one sound-file is extremely corrugated, and we apply the time-varying amplitude envelope of that sound to a fairly constant-amplitude sound-file, then we can end up with a second corrugated sound-type.

Window length: The length of the 'search window'. The smaller this value, the finer the resolution is (picking up short-term fluctuations in amplitude rather than larger overall shapes). Note, however, that by making this value too small (ie a few samples) you will end up seriously distorting the second sound.
Normalise before enveloping: You have the option to try a force the envelope of the second sound up to 'maximum gain' before applying the calculated envelope to it. The advantage of this method is that the final envelope more closely matches the original envelope; however, the drawback is that background noise can get amplified to maximum volume - a Bad Thing...