There are a lot of very good ML-based tools for improving audio that are able to do things that were previously impossible. Might be worth a shot. (I haven't used this one, just the first that Google popped up: https://audo.ai/)