No, he is demonstrating that digitization (sampling) of any audio waveform must be reproduced as the same waveform, because there is only one possible waveform that can fit the sampled points. This is true regardless of the filetype/compression used, though at very low sampling rates (below 8 bits/second) information is noticeably lost because there aren't enough sample points in a given timeframe, and then you get distortion during playback.

I'm not sure specifically where the line is, but there are physical limits to how fast a speaker cone can vibrate to reproduce the waveform no matter how detailed it is (it would depend on the specific playback device), regardless of whether the source is analog or digital. There's a point where the level of detail of the sampled waveform doesn't matter, because there's no audio equipment in existence that can reproduce that waveform with any accuracy, and 320 kHz is definitely above that point.

It is more likely that the "muddy" quality of the MP3 files you're referring to is related to choices made while converting/compressing the file, such as the differences in dithering that he demonstrates later in the video.

[–] BCsven@lemmy.ca 1 points 6 days ago* (last edited 6 days ago)

Nope, mp3 can still be 44kh sample rate, but compression algorithm loses fidelity. Its not 100 % reproducible. This the name lossy compression vs lossless.

Edit To be clear: he's shown live DAC, not extraction from compressed audio. To illustrate that we can capture the wave and reproduce...but when you compress with lossy, its lost.

There is a website where a person put mp3 output back through as input multiple times till the sound gets garbled through entropy. Couldn't find the link, but this dude did the opposite and captures what the mp3 algorithm drops. So when you listen to it it is mostly quiet but a lot of sounds left behind. https://www.factmag.com/2015/02/19/listen-song-made-entirely-data-lost-compressing-suzanne-vegas-toms-diner-mp3/