I was tasked with an audio restoration project this week and was reminded that I have never felt the need for a de-Esser when I felt the De-Essers I own, have used, or demo tested seemed to fulfill the requirements.
In my experience, if you really need a de-esser, none of them do a good job.
The only time I have found a de-esser to seem effective is when I had become unduly fixated on sibilance that was not actually problematic or objectionable.
The only way I have found to satisfactorily tame out of bounds essing is with hand edits in a spectral editor. I use iZotope RX for this.
For this recent project, I worked with several hours of a poorly recorded oral history interview sourced from a deteriorated compact cassette. Because of the length of the content, I did not go down the hand-editing rabbit hole. I did try some of the brand name de-esser plugins I own licenses for but eventually abandoned the hope that I would find them useful this time.
This recent experience made me think that someday somebody is going to replicate the decisive precision of hand editing spectral content with some sort of machine learning artificial intelligence automated technology.
Has it been done? Did I miss it?
I am just throwing the idea out there in the hopes that it might spark an interest.
Thank you!
