What's out there for neural network based audio restoration?

I have some 42 year old "samples" I made for my EH Instant Replay on cassette. They've been in a nice case all this time. I went around with a professional quality recorder and mic back in the day.

I'm aware of the noise reduction in Audacity, and Steinberg's "Clean".

Looking for a neural based solution, hopefully recent.

 

You probably want Steinberg Spectralayers or iZotope RX. I prefer Spectralayers because it has ARA support, so I can work directly inside the DAW. I've done lots of work for an upcoming movie using it. Excellent results.

 

Waves Clarity Vx Pro - www.waves.com/plugins/clarity-vx-pro
Neural Analog - https://neuralanalog.com/restore-audio-mp3-to-wav

 

You should check around "audio super resolution" or "audio upsampling" terms.
UVR can now accept denoising models also, but I overlooked this a long time ago and I won't be able to provide precise instructions.

Concerning making the content HD by "cleaning" the high end :

There is a ton of open source projects, I had a quite good list but not at reach right now. I was initially looking of an "Adobe Podcast" local alternative. There are even denoising models around.

You do have this one : https://github.com/haoheliu/versatile_audio_super_resolution

They advice to low pass content in order to have a stable cut off frequency, especially for jittery cutoff from lossy codecs. But if it's tape high frequency dampening this should be very fine. Produces 48Khz PCM WAV output.