You should hear the entire phrase "continuous soft and relaxing", intelligibly, going forward. You might want the backward part to go by a bit faster than the forward part. The result should sound something like this.
The patch should load the sample into a wavetable using "loadbang" and "soundfiler" objects (as in the documentation patch B07.sampler.pd), but make sure the wavetable in this case is big enough to hold the entire soundfile. To keep the patch itself reasonably small, clear the "save contents" property for the array.
Although you might later have to include soundfiles with your homework submissions, for this time around you should just assume there's a copy of "voice.wav" in the same directory as the patch so you can just invoke the filename "voice.wav" when opening the file.
At this point, you should always be making a nice, de-clicked output level control
in decibels, as in the examples for this week.
Depending on how you carry this out, the sample output might come out either
clean or dirty. (The sound clip I made went the dirty route.) Either
way is acceptable.
Sorry, no extra credit suggestion for this week.