Help with audio restoration

[DaneClark]

I like to collect episodes of old cartoons that aren't commercially available on DVD and restore them as much as possible to make my own personal DVDs. I have a bunch of hard-to-find episodes of Alvin and the Chipmunks that came from a multi-generational tape with very muffled sound. I don't want it to sound perfect, I just want to make the characters voices easier to understand. I have soundforge and audacity, and I have also provided a sound clip from one of the episodes so people can hear it and help give me some advices on how to make it sound clearer.

[transporterfan]

This is about as involved as it gets.
Apart from major hiss removal you have a rich base from the human characters, a high treble in the chipmunks voices and a residual sibilance to all the speaking parts. The last one can probably be reduced but I doubt if it can be removed entirely.

The human speaking parts are relatively simple to isolate, but listening to it 'raw' the chipmunks will prove almost impossible to enhance.
The added noise from multi-generational copies adds not only noise, but also produces a smoothing effect overall, particularly in the treble. Removing noise will most likely produce more distortion.

But I like a challenge. I'll run it through a parametric EQ and try some noise reduction filters.
There will likely be some mixing of tracks too (base, treble and mid seperations).

I will get around to it eventually and let you know what comes out. Keep checking back.

[transporterfan]

Each of your tapes will be different but the principle is the same. I have not used SoundForge so don't know it's capabilities, but in Audacity use Google to explore how to use the noise reduction filter: find a 'quiet' area (no music, no speaking parts) and copy a section to the clipboard. Use it as a template to show Audacity what 'noise' you want to remove. Once that is reduced to your liking use an equaliser to adjust the frequencies. I imagine SoundForge has both the requisite tools to accomplish the task on its own. I used GoldWave.

As you can hear the sound is clearer but much of the information has been lost through the various copying processes. It is impossible to restore what has been damaged.

http://files.videohelp.com/u/198336/Alvin__UA_Filtered.wav

[Mephesto]

Transporter fan, the OP was asking how to mitigate the muffling (dullness) of the audio not how to denoise which is trivial and hardly even necessary for this clip.

OP, there is literally nothing except noise above 7khz in your sample. You will never restore the missing frequencies and your sound will remain muffled unless you can find a higher quality source of whatever movie you got. All you can do right now is delete everything from 7-48khz to get rid of the noise. This may or may not sound more pleasing.

What you can do is resample it to 7khz then back to 48khz with no anti-aliasing and this will unintentionally add some upper frequencies from already-existing lower shelf which again may either add some texture or sound less muffled but more annoying.

[2Bdecided]

I think the limit of the real content is about 3.5kHz.

There are "high frequency re-birth" type filters to fake high frequency content, but working from a 3.5kHz bandwidth signal is probably pushing them too far.

Doing Mephesto's trick will intentionally add aliasing and make the audio sound really metallic (like a baby's toy with a cheap sound chip). Might be OK for the chipmunks, but probably not the rest of it.

[netmask56]

I doubt if there are any 'domestic' plugin exciters that would recreate faux harmonics above 3.5Khz with any acceptable aural effect.
Given access to one of these is another story http://www.aphex.com/aphex-products/exciter/

unlike an EQ, the Aural Exciter’s harmonic processing doesn’t boost the signal’s output level, maintaining a clear and balanced sound. Traditional EQ increases the output level of the targeted frequencies. Adding EQ to boost selected frequencies typically results in a perceived loss of other frequencies, which are then also boosted to compensate. The Aural Exciter’s patented technology harmonically brings out the desired frequencies without adding gain. The result is audio that’s more balanced, more articulate and, simply put, better sounding.

[Cornucopia]

An exciter is just a specialty (usually band-limited) harmonizer, so you can combine plugins to achieve the same effect if you know what you're doing. And BTW, yes you can use material that is band limited to 3.5kHz, you just have to know what subset of that material you should use, and how much (think: last octave's worth + 1 octave pitch-change/harmonizer w/ feedback, also think: subtle).
Resampling won't work, for the clear reason that 2Bdecided mentioned.
Also, contrary to Mephesto's claim, one the main, important & first steps in "restoration" is isolation & noise reduction.

Below is a sample, similar to TF's, yet dissimilar. There is less "harmonic grating" and overblown bass. But then, I overdid it with the LF vs. HF balancing (took out too much LF) and overdid it with the Broadband NR (works, but you can hear birdies). I also used harmonic notch filters, which I think helped. Plus, I did a mockup example exciter (only added it at -48bB, but I think it is a little clearer) like I was talking about. Used BP of 1.6kHz-3.2kHz and pitch doubled to 3.2-6.4kHz. All of it was Quick-n-Dirty (~20min of operation) as I don't have time to do it right, right now. I'm working on Sound Design & Graphics for H2$.

Scott

[transporterfan]

I like that Scott, a lot. Very good. I look forward to listening to the finished example.

[Mephesto]

Also, contrary to Mephesto's claim, one the main, important & first steps in "restoration" is isolation & noise reduction.

You misread, I said for this clip all you have to do is delete everything above 3.5 kHz to get rid of all the noise. 99% of legit content is below 3kHz actually and 90% below 2.4. There is noise in all bands but it's largely inaudible in the legit shelf. So everything above 3kHz (6Khz in terms of sample rate) can be safely deleted and all the noise is gone.

Below is a sample, similar to TF's, yet dissimilar. There is less "harmonic grating" and overblown bass. But then, I overdid it with the LF vs. HF balancing (took out too much LF) and overdid it with the Broadband NR (works, but you can hear birdies). I also used harmonic notch filters, which I think helped. Plus, I did a mockup example exciter (only added it at -48bB, but I think it is a little clearer) like I was talking about. Used BP of 1.6kHz-3.2kHz and pitch doubled to 3.2-6.4kHz. All of it was Quick-n-Dirty (~20min of operation) as I don't have time to do it right, right now. I'm working on Sound Design & Graphics for H2$.

Scott

Everything you're saying sounds all cool and everything but I looked at the spectrograph of your sample and I see no restoration of frequencies. None. Just noise reduction.

You could transpose and then mix the top shelf of the transposition with the original but this sounds really unnatural, annoying and would work better for a 22to44 restore not a 6to12.

[Cornucopia]

Except one can easily do a 18dB/Octave HPF @3.5kHz or even 4kHz and still hear things, so clearly there is material there NOT just noise.

Also, didn't you notice where I said 1. this was quick-n-dirty? and 2. I only added back in at -48dB? It's not likely it would show up much on a spectrograph. The reason I only added a little was because I whipped it up quickly, so it had other crap still in there which you wouldn't want to have added. Any kind of transposition sounds annoying if you use too much of it, even if you are using single instruments like guitar (where this is a commonly used feature, e.g. Santana). If you just apply a small amount it is enough to "open", "lighten" or "freshen" the sound.

Of course, the aforementioned idea of using a better source is to be preferred. Lots of this restoration happens, though, just because a better source is not available. And "you cannot make a silk purse out of a sow's ear". But you can improve this material in a number of ways, and NOT just by LPF.

Scott

<edit>TF & OP: I will be prepping for H2$ which opens in 1 1/2 weeks, and I'm building a new demo reel for an important position that I'm applying for, so will be unable to get back to this with any regularity until ~2 1/2 weeks from now. That clip was mainly "example" and "proof of concept" anyway.

[Rugratskid]

I like to collect episodes of old cartoons that aren't commercially available on DVD and restore them as much as possible to make my own personal DVDs. I have a bunch of hard-to-find episodes of Alvin and the Chipmunks that came from a multi-generational tape with very muffled sound. I don't want it to sound perfect, I just want to make the characters voices easier to understand. I have soundforge and audacity, and I have also provided a sound clip from one of the episodes so people can hear it and help give me some advices on how to make it sound clearer.

Can you please please please send me some of the episodes somehow, muffled or not? Some of them like this one, aren't even online yet!

[NerdyNette]

Thank you SO much for posting this!

Do you have any other rare episodes of The Chipmunks? The quality does not matter, as if it's rare, it's better than what we have!

I'm a huge Alvin and The Chipmunks fan, and the episode you've posted the audio for was previously not available in English at all.

[lordsmurf]

I have all of them in A+ quality. (Find me at www.tvpast.org to discuss that.)
Getting back to the tech topic at hand...

Attached is a quickie version. I don't want to spend a whole lots of time on it. It was done in Audacity and Sound Forge. The better option is to find a new source. I only have eps 1-65 here. I'm not sure where 66+ are right now, otherwise I'd attach that too.

I ran these digitalFAQ.com SF filter presets, in this order:
Audio Restoration 5 ~ tweak 70db
High Restore 13
High Clipping 3
Volume 350%
High Clipping 3
Remve Bass Grubmle 4
High Clipping 1

For this clip, I just started with what Cornucopia had already done (shortcut for myself). This could be done several ways.

Just a quick sample of what can be done...

[Mephesto]

Lordsmurf, before I bother looking at your restoration tell me if you worked on the original sample provided by the OP or another source you found elsewhere because I'm not interested if you did the latter or overlaid another source on the OP's.

[Jay55]

Due to price this probably isn't practical but wanted to put a word in for Zynaptiq's new Unfilter VST plugin. It's quite a novel tool for restoring missing audio data. It's only been out for a few months and there's nothing else like it (Developed by two guys from Hanover Germany). Demo avail: http://www.zynaptiq.com/unfilter/ - Youtube example of what it can do: http://www.youtube.com/watch?v=7thIB9VX35k

[Dogway]

Unfilter seems to be an automatic EQ, an auto frequency based gain flattener or kind of, so for this case <3KHz it's not gonna do much I fear.
I have a question though, would an exciter add richness (harmonics) to existing harmonics, or generate new HF harmonics/content pass the cut-off based on mid freq?

[mammo1789]

Izotope RX has some nice filters for restoring maybe you can try that

[Deter]

Izotope RX costs $349 and for advanced about $1,200. With the demo program u can't save files.

Wanted to test out Protools, just trying the demo was a nightmare, You must have a second-generation USB iLok.