That seems pretty improbable. While the CPU may be fast enough, the real problem is that it's nearly impossible to identify sound types that should universally be ignored. The feature would be special case, and would require several profiles, one per "environment" such as recording different types of music, or a voice performance of some sort, or other things. Also, frankly, I'm somewhat against the idea of attempting to bloat the core recording code with features that are generally designed specifically for "stealth" recording, which seems shady at the least for a project that generally likes to make clear how much we respect licenses.