it's gone in a single still frame

but across many consecutive frames, the information is spread out temporaly and can be recovered (partially)

the same principle of how you can get a high resolution image from a short video, by extracting the same patch from multiple frames

https://en.wikipedia.org/wiki/Video_super-resolution

That looks interesting. Is there ready-made software that can do this? Doesn't have to be easy to use just useable with a time commitment of a few days.

No, it's not "restoring detail". The information is gone.

It is predicting what the information might maybe have been like.