This is fine. I'm able to restore speech. Now a non-speech recording fixer could be really useful here. Any hints on how to train my own one on that?
