Detecting deep-fake audio through vocal tract reconstruction: Difference between revisions