It’s difficult to validate strings that are of so different volume levels as those are that we’re asked to validate. Sometimes, I can barely hear anything, sometimes, the first syllables hurt, until I can lower the volume.
Could some normalizer be applied to the sounds before you ask people to listen to them? Maybe even in Javascript, so your original data doesn’t need to be changed and the work is done by the client browsers?
I’m not going to continue validating until there is some normalization in place, it’s just very uncomfortable.