Would it be possible to somehow “fake” the high-end on a low quality recording (either downsampled from 44100 or just a low quality live concert recording in my case)? I know neural networks are being used for enhancing pixelated pictures; could the same thing be done for audio? Have it make educated guesses as to what the high-end sounded like, or are there tricks which can be done which will “fake” the high-end? An example of the quality I’m talking about is here: Listen to a few seconds of it and you’ll hear it’s very low in quality (both audio and video)
Vocal exciters, (a/k/a “Aural Exciter”), can add high missing high-frequencies, but they are only usable on clean, (not crackly), recordings of the voice /instrument in-isolation.