I'm not buying it. The sound recording technologies of the late 1920s did not have a wide frequency response, a high dynamic range or such a low noise floor. There is no drop out typical of early recordings and there is no discernable wow and flutter. Without precise mechanical timing in the recording mechanism, the pitch and tempo would be all over the place since the speed of the recording would not be stable for proper playback. The audio has too many red flags to just accept as real.
Yeah, i've watched a bunch of these from different uploaders.
They just take audio from an online audio file archive and adjust speed , pitch , tempo etc to replicate what they think matches as best they can.
Same as the bs colorization AI, simply guesswork and basically faking what the actual colors, shades and textures are.
Take a photo of something with multiple colors (a large book case with dozens of different colored books ) or a scene with several different shades of different color, make it black and white and run it through an online colorizer app a number of times. Which one is the correct one - if you didn't already have the original ??
I have no idea what the author did technically to resolve these audio issues.. You could ask him how he overcame these issues in the comments on the video.
(post is archived)