As an enthusiast of sound recordings from the 1910s and 1920s I’d like to see some of this technology applied to resurrecting early performances. Much has been done to clean and upgrade old recordings but the results are limited by the narrow frequency response of early recording equipment, especially those of the pre-electric era. I wonder if AI techniques can be used to restore–or rather, recreate–the full range of sound based on knowledge of the instruments used and the recording environment. It’d be wonderful to hear vintage world-class orchestras, early jazz and dance bands, or pioneer blues musicians as if they had been recorded today.
In the best of all possible worlds these would be open-source tools so that old music, already plagued by copyights that never die, wouldn’t be further locked up in the private rentosphere. But that’s another story.
