Help me fix my audio degradation please!

[tribal nymph]

Recently I've developed an obsession with the public access tv show rAw TiMe and i have some videos with severely degraded audio, if i could get any assistance with restoring it to a somewhat acceptable state, that would be extremely helpful!
The videos I'm using (while not all of them have degraded audio) can be found here: https://archive.org/details/june-26-1999-tiffy-and-george-marshall
and here: https://archive.org/details/rawtime90
the affected videos are: August 14
august 21
and july 10

[ruby sonnet]

You could have a look at these, might be you're gonna find something useful to help your case:
https://github.com/resemble-ai/resemble-enhance
https://github.com/haoheliu/versatile_audio_super_resolution
https://github.com/haoheliu/voicefixer
https://github.com/JusperLee/Apollo

[shut mantle]

dango is rather paywalled but it can enhance better than apollo in cases like this, though may not reconstruct harmonics better than audiosr (I suppose?)

[ruby sonnet]

Haven't tested Dango or Apollo personally, but did the vasr, it's quite alright when you get the settings right ( else it's mostly hit or miss ) yet biggest issue I subjectively have with it is the inconsistency, sample:sample wise inconsistency

[tribal nymph]

i appreciate the suggestions , but unfortunately i don't have the proper technological skills to be able to actually use any of these

[ruby sonnet]

https://huggingface.co/spaces/ResembleAI/resemble-enhance

Here's some hugging space you can try

( they run it for you )
Just unsure of the limitations as I barely use such

Worth giving a shot

The interface should be pretty straightforward

You'd just have to find some setting that does the trick for you ( if even

[tribal nymph]

unfortunately there's a one minute limit on audio, which is around 89 minutes too short

[ruby sonnet]

You can always just trim your files

and do piece by piece

That's unfortunately how it typically is

Can't imagine doing complex inferences on such long files ( most of the time

[shut mantle]

im sure there are some colab notebooks you can find

[ruby sonnet]

Or that ^ But I ain't aware of any, sadly

[shut mantle]

I know only colab for apollo as well as audiosr

[ruby sonnet]

Then you could try those 2 ^

( and even the 1 min limit one, for demo / to see if it's even worth the hassle

[tribal nymph]

if you'd be able to give those to me, i'd be appreciative!

[shut mantle]

https://github.com/jarredou/Apollo-Colab-Inference/
https://github.com/jarredou/AudioSR-Colab-Fork
https://github.com/jarredou/hifi-gan-bwe

[tribal nymph]

thank you

[tribal nymph]

unaware if im doing something wrong, but like, im getting an error even with a minute long clip, am i doing something wrong, or what

ive changed the settings, kept them the same, error every time

[ruby sonnet]

What does the error say?

or just " error "

You sure it's exactly 1 minute or, ideally, under it?

even half a second or 100 ms could cause an issue
( In cases where there's a hard-limit on input length

[ruby sonnet]

ahhh, well, perhaps the space is broken then

In that case colabs is your way

But I think, makidanyee could help you better

[shut mantle]

idk but perhaps out of memory error?
the colabs I showed you above can process the audio at any length using chunk & overlap method