Our music workshop will include a listening exercise using several generated exports. I want the discussion to focus on sound rather than assumptions about software. I would label the excerpts with neutral letters and match their playback levels. Each excerpt would contain the same musical passage to keep comparisons focused. I would ask for timestamps and descriptions before accepting an explanation of the sound. The group includes people with different listening experience, so technical vocabulary should be optional. I would collect individual notes before discussing them together, and let students replay the excerpt when they need to check whether their first impression holds up. I found a resource on ai audio artifact removal that might inform an optional later comparison. It has not been tested for this exercise, so I cannot claim a benefit. The exercise needs room for the original version to be preferred. Which simple criteria would encourage precise observations without steering the verdict?