Subject:Need a clever declicking/repair tool

Posted by: Doug_Marshall

Date:1/11/2014 3:59:30 PM

I've got a wave repair problem that I really don't want to give up on. Below is a link to a short file in which you can hear a cyclical clicking caused by some sort of sync(?) problem during the recording session. The result is that the waveforms for most of a day's work have this problem. If you analyze the click you'll see that it is is caused by brief missing chunks (and time) in the waveform.

Does anyone know a tool that could either reconstruct or repair such data? I know there is software that can automatically find perfect loops for sound samples. I would think this would be analytically similar, though a non-looped goal.

All help appreciated! Below is a link to a short snip of the file:

Doug

https://dl.dropboxusercontent.com/u/3729273/ClickyFile.wav

Subject:RE: Need a clever declicking/repair tool

Reply by: Kit

Date:1/11/2014 5:30:39 PM

I would imagine that izotope's RX3 would handle that but it's an expensive solution.

http://www.izotope.com/products/audio/rx/

Subject:RE: Need a clever declicking/repair tool

Reply by: Doug_Marshall

Date:1/11/2014 6:21:19 PM

If RX3 can do it, it's worth the price! But I'll try to pose the question to them first or hope there's a demo that's adequately functional. Thanks!

Subject:RE: Need a clever declicking/repair tool

Reply by: Kit

Date:1/11/2014 10:04:19 PM

They have a fully functioning demo (apart from save) so you could give that a go.

Subject:RE: Need a clever declicking/repair tool

Reply by: ChristoC

Date:1/12/2014 2:14:22 PM

Wow, that's a tough nut to crack! For sure a clocking problem..... I'm very limited for time now, so I just had a brief look - used Cedar DeClick to fix the discontinuity in waveform at each click, but that still leaves artifacts; couldn't find a 'quick' solution with RX3 either..... maybe you have better luck by spending some time. The clicks during the reverb are going to be the trickiest....

Subject:RE: Need a clever declicking/repair tool

Reply by: Doug_Marshall

Date:1/14/2014 5:29:18 PM

Thanks for giving it a go, ChristoC! I'm probably wrong, but it seems like it should be possible for a software tool, if one exists, to piece things together just like a looping tool does with samples. I have a long defunct sample looper (Seamless Looper) that typically finds dozens of perfectly tick-free loops in almost any steady-toned file. I have made a few of these repairs manually. It takes insane amounts of time to piece things back together by eye/ear but it can be done. It's just that nobody has the time to do it with, say, a thousand files by hand...

Subject:RE: Need a clever declicking/repair tool

Reply by: MarkWWW

Date:1/15/2014 2:02:09 PM

Yes, this file has regular sections of data missing from it, presumably because of a sync mismatch, as you say. But I believe that the regularity may be a help in repairing it (though it won't be perfect).

As far as I can see the problem with this file is that it has regular chunks of 22 samples missing every 24000 samples (i.e. every quarter of a second), starting at sample 10751. (My guess is that receiving clock was running 88 Hz slower than the source clock and was re-syncing 4 times per second, causing 22 samples to be lost each time it resynced (re-sunc?).)

By modifying one of the sample SF scripts (Add Markers) I have produced a script that will insert a marker every 24000 samples, starting at sample 10751, but at that point my programming expertise runs out. If someone (not me, I'm afraid I don't have the programming skills for it) were to write another script that would insert 22 samples (zeros would probably be best) after each marker then the file would be restored to the correct length and this would also put the various sections of the file into the correct time/phase relationship with each other. Once the file is restored to this condition it should be relatively easy for a spectral repair tool like iZotopeRX3 to cover the empty gaps with plausible sounding material taken from each side of the gaps. (It's the time/phase issue caused by the missing samples that is making it hard for iZotopeRX3 to do a good job with the file as it stands, I think.)

If all your problem files have the same characteristics as the sample you posted then it should be possible to rescue them in a similar way, I think.

Best of luck

Mark

Subject:RE: Need a clever declicking/repair tool

Reply by: Doug_Marshall

Date:1/15/2014 10:41:09 PM

Hey, thanks for that insight! Since all of these files consist of individual notes that will ultimately run in independent loops when called, the missing time in the waveforms doesn't need to be restored, only the continuity of the wave. When I repaired a few notes manually, I simply worked to find zero crossings that matched and pieced them together so no tick was generated. It was agonizing, but it did work. I didn't necessarily attempt to repair the whole length of every note, just enough of it to make a musically reasonable sample loop.

As I mentioned before, it seems to me like a challenge very similar to that of finding seemless loop points for samples. There are automated tools for that. How I wish there were for this problem!

Needless to say, I learned my lesson to very carefully check the actual recorded results I'm getting, early and often.