Just around a century ago, actors started speaking in movies. Movie stars like Al Jolson, Dolores Costello, Conrad Nagel, Ronald Colman, John Barrymore and Greta Garbo suddenly had voices.
But many of those old movies are low quality by today’s standards. Between transfer problems, degradation over time, and simply the limitations of technology, the sound and images are very different from what we would expect today.
On the one hand, it would be tempting to use A.I. to upgrade them. On the other hand, I wonder whether that upgraded version would truly be faithful to the original performance.
Using an AI model to modify a performance necessarily relies on training from many other human performances. The “higher fidelity Jolson” might not actually be entirely Jolson, but rather an amalgam drawn from a variety of sources.
Maybe its best to simply watch those old films in their original low fidelity condition. At least then we will know that when Garbo speaks, it will be 100% Garbo.
I have seen remastered videos of the streets of Vienna in the early 1900s, and the added color has a way of making the scenes more … relatable maybe?
But I also wouldn’t want old movies to be changed, or texts of old books, etc.
The exception being the Austrian anthem: after a lot of discussion, the original lyrics were changed : “home of great sons” was changed to “home of great daughters and sons”.