Digital clipping in original audio source...How 2 fix?

[Moby Ball]

I have Goldwave, Cool Edit, Sound Forge, Cubase and Sonic Foundry Acid. I've used the various EQ and Noise Reduction options provided by these programs but I just can't seem to find the right section of bandwidth to knock-out the digital clipping (distortion) without mangling the overall sound of the track.

Is there any way to get at the problem areas with out doing damage to the rest of the psychoacoustic space?

Cutting out hiss or hum is easy, but I can't seem to find the right levels to get rid of the clipping. Does anyone have any specific level numbers I could try out?...

I have an alternate audio source that sound markedly better, but I can't get the sucker to sync to the video with the problem audio source. They are two different lengths, and two different bitrates. I've tried using the time-stretch feature in SoundForge, but for whatever reason, the audio is always over-stretched by 40 to 60/1000ths of a second. Also, the stretching doesn't seem to be even, across the length of the audiostream. The files start in sync, but within seconds, they are grossly out of sync. I would try the old VirtualDub tricks to fix it, but the files are both svcd - one from a dvdrip, the other from satellite - and my version won't handle mpeg2.

If anyone can give me a hand on either means of going about fixing the audio, I would appreciate it very much.

Thanks,


- Q

[a]

Oh well, if you're really prepared to go through with this..

Sounds like your original source has either been a slightly streched VHS tape, or there is some other similar problem. In cases like these you can't work with the whole length of the soundtrack, but need to cut it into pieces. It is impossible, repeat, impossible to fix a clipping digital audio track in any other way than cutting away the clips and trying to attach the soundwave together in as unnoticeable manner as possible. With a video source, of course, you should be prepared to either stretch or cut a part the video too to match the amount of frames you removed from the audio.

If you have an alternate audio source, you're slightly luckier. I expect you to be familiar with Sound Forge, and in any case you're going to work for a long while, since you need to do it near professional style. There is only one more possibility: count the length of the video and audio, it is possible that the both sources are different standard (the video on NTSC and your external audio track in PAL, for example). This would make it possible to convert the video only to different standard, and then multiplex the audio to it. If lucky, you're home free. If not:

You have three sources. Original video (A), the clipping original audio (B), and the better quality external audio (C).

a) Get TMPGEnc. Use the demultiplex function to divide A and B from eachother. You should also check, which encoder and codec was used to compress the audio and get the same programs, if possible.

b) Use an mp3 encoder to create a constant-bitrate WAV file out of B. If you couldn't get the proggies used originally, I prefer (Razor)Lame myself.

c) Fire up Sound Forge, open B in a new window. Start dividing B into parts using markers, depending on how you hear sounds and sets change. Open up C in a new window and a third window to store "final audio".

d) Copy the markers from B to C, now you got two audio sources very closely, but not quite, resembling eachother.

e) Wish to die? Great.
e1) _Mix_ the first part of C into B. Check the waveform and listen closely.
e2) Didn't match? Undo, timestretch or edit the first part and start from e1.
e3) Matched? Cool, Undo, Cut and insert the first part of C into the third window. Delete the first part from B.
e4) Now you have two audio sources which are slightly shorter, and a new audio source which is slightly longer. Start again from e1 until you have only the third audio source left, let us call it "C2" from now on. Do a finel check, that the audio track truly has the exact same amount of frames as your video source, and save the WAV file.

f) Use your favourite encoder to create a compressed MP2 file out of C2, for best results it is advisable to use the same format, compression ratio etc as B before you uncompressed it. DO NOT DELETE THE ORIGINAL C2!

g) Multiplex A and the new compressed audio track of C2 together in TMPGEnc. Do a test view and hope for everything to be in order. If not, make a note of the point, where you lost it, fire up Sound Forge again and redo the whole process from the point it started going to the wrong direction. Eventually you will have a perfect audio track, which you can mux with the video stream and then start bragging to your friends about it.

Too bad for you, unfortunately there are very little automated procedures to bypass any of the steps I wrote about here. Happy projects and remember, no one's paying you for it :p

[Moby Ball]

Thanks for the reply...

...Unfortunately, I've already tried all the methods you suggested here. Perhaps I just need to execute them better...I'm not entirely sure where I'm going wrong with the files, but I'll give it another shot. I just wish the damn thing wasn't so time consuming: I've already put a good 21 hours into the audio fix alone.

Much Appreciated,

- Q

[a]

If the video and audio source are exactly the same length, I find it strange for the process not to succeed. All I can think of is your source being randomly stretched, so that your playback speed actually goes unnoticeably up and down like a wave.

In this case 21 hours might not be nearly enough, you need to divide the audio track into minimal sections, possible less than one minute each, to get succesful results with minimal time stretching, cleaning, editing, doing final mixdown, EQ, possibly even compression during the mastering process. This is pretty much the poor man's professional restoration procedure, so there's no other way to go if you want the fix to be near exact.

Since you need sharp ears and have to concentrate to compare the waveforms, it is possible you have faced what the audio industry calls "being deaf to the music". Do only a few minutes (let's say 3 blocks one minute each) worth of the final audio track at a time, comparing the waveforms and stretching/editing where necessary, and then take at least half an hour break. This prevents you from losing your touch and getting the sounds garbled up in your head, so that you won't notice the tracks are indeed out of synch even if they look like they are synched. It's also a good idea to play the final audio track on top of the original extracted audio after every 10 minutes of finished up work, so you can tell if the minor differences (which there always are) are truly unnoticeable or not.

Unfortunately, this is not anything user-friendly, but you have to start from the beginning every time. First listen to 10 minutes of the mix, then 20, then 30, all the way until you're at the very end of the, say, 95 minute movie. You're really having a tedious task at hand. Like I said, at least you can brag about it afterwards :p

[Moby Ball]

If the video and audio source are exactly the same length, I find it strange for the process not to succeed.

Ah...an therein lies the rub. The audio sources are not the same length. The dvd source audio is 38 seconds (roughly) longer than the satellite audio source at the beginning, and 93/100ths of a second shorter at the end of the "sound" ("sound" = dialogue and music...prior & following to which is just dead-air space).

To deal with this, I've been eyeballing the original audio in SoundForge, and marking where the "sound" begins. I write down that time and then cut the files beginning off to the point where the audio begins. I then do the same thing with the higher quality external audio source, and cut it down to the point where the "sound" begins. Thus, I have both files starting at the same place.

Then, I find where both of the files end. I time stretch the audio on the external source to match that of the original problem source. Following that I insert 38 seconds of silence at the beginning of the external source to match that of the original. Then I add the extra few thousandths of a second (5/1000ths I think) to the end of the external source so that to all appearances the files are matched.

HOWEVER...they aren't.

The sound holds together for about 3 seconds and then falls apart.

So...I suppose that I will have to do as you have said, and chop the file into 51, one minute increments and sync them individually, however, my question is whether or not I have done the right thing in chopping out dead-air space at the begining of the audio files, prior to making the time stretch adjustments?

All I can think of is your source being randomly stretched, so that your playback speed actually goes unnoticeably up and down like a wave.

I've wondered that myself. It doesn't seem to follow a set pattern. But I'm going to start again from scratch and see what happens.

As for getting a dead ear...

I know of the phenomenon, but I rarely, if ever, have problems with rhythm or meter. I'm a jazz bassist and dj and do alot of beat matching so I'm somewhat attuned such discrepencies. Where I have difficulty is in picking out certain frequencies, hence, I let most of EQing responsibilities fall to my partner who works with me in our studio. This is why in my initial post I was asking for help with getting a specific EQ setting for the problem audio source. Thanks for the consideration though...it can get exhausting listening to the same snippet over and over again.

L8ter,

- Q[/quote]

[a]

So...I suppose that I will have to do as you have said, and chop the file into 51, one minute increments and sync them individually, however, my question is whether or not I have done the right thing in chopping out dead-air space at the begining of the audio files, prior to making the time stretch adjustments?

Sure, as long as you get the sound starting at the same location and synched to the video you have no problem here, the rest is only tedious fine-tuning. Cutting or adding soundless parts to the beginning or end of the file is completely safe, as long as you get it otherwise synched.

Sounds like your satellite source might be 30fps and your video source 29,97fps among other minor differences (for example the TV station using an old film reel or (S)VHS cassette for broadcast). At least these possibilities seem most likely. I'd recommend using max 1 minute parts for editing, time streches on this scale are pretty much unnoticeable in the final mix, anyway.

I have difficulty is in picking out certain frequencies, hence, I let most of EQing responsibilities fall to my partner who works with me in our studio. This is why in my initial post I was asking for help with getting a specific EQ setting for the problem audio source.

Well, can't help you there as long as there is none :p You could easilly use EQ settings to "muff up" an audio track with analogue saturation, but this digital "harsh" overdrive is pretty much impossible to fix so that you wouldn't notice it. Your best bet here would be to cut away each pop one by one matching waveforms, but then you would again be in the same position you are now: you have a video and audio source which are not synched anymore.

What you could do, of course, is trying to cut away each frame with a pop. If you manage to make it unnoticeable, you could always stretch, say, 4 frames around the cut by 25% to cover for the lost frame. It's a long shot, but you might get lucky using this procedure. After removing the clipping frames i'm not sure how bad the audio is, the basic preparation process in Sound Forge would be using a sample from a silent part of the audio track to create an "image" of the background noise, noisefilter the whole track using this sample, and then slightly boost the low frequencies, and perhaps little less the middle ones to patch things up again. Again, if you get lucky you might get satisfactory results, unfortunately these procedures are always of the trial and error kind.

Like with music, EQing is usually first done with multiple tracks separately, and only minor nyance changes are worked on the audio after the mixdown is already complete. Removing background noise or EQing a flat audio track for a video means having to deal with so many frequencies at the same time, your best result would be a very metallic and thin audio track which would not be nice to hear - and still very likely to include most of those pops you want to get rid of.

Repairing damaged audio track alone is a completely different process compared to working with video and audio together, this time you cannot just fool around with nyance or atmosphere but also need to be exact by the millisecond to get things going on. Sorry I don't have any shortcuts for you to use, if timestretching the whole audio track didn't work, it's either this choice or you need another video source.