Not that I'm not capable of creating a crapping recording, but I must confess I had nothing to do with this one... just trying to do something nice for the artist.

I also discovered the hidden crap I hid using the smoothing function. The wav forms look nice after you perform the smoothing.. you actually see reasonable peaks instead of that straight line. So I was surprized that the crap really wasn't gone at all. I thought I could pull out what good highs were left after the smoothing.

Such a pretty song though, maybe it could be useful as a back background track with some ambient sound to cover up the crud.

I suspect that if we could readily identify distortion visually in the waveform then it would be a lot easier to remove. The problem is that it is extremely difficult to identify what portion of the waveform is distorted and what isn't. About the only method we have to reliably identify it is with our ears. Other than simple things like clipping, because sound is so complex software just can't distinguish between distortion and what's really supposed to be there.

I was once handed a recording to be used for a dance show that was so badly distorted that it was almost impossible to make out the melody. It was an 80 year old disc, out of print, and no chance to find another copy of it before the performance that evening. I took the recording to my brother, who is a professional musician and had him listen to it a few times. He was able to piece together enough of the melody by ear to play it on his keyboard and produce a new recording for me. This is the sort of thing that software just can't do.

The "trick" is in the EQ. Every sound pro knows that you fix things by taking away what you don't want. Like feedback in a microphone to a speaker. The engineer doesn't add more bass to take away the treble, he takes away volume first, then takes away treble, then adds what he needs back into the mix. This is different than painting a picture. You put in what you want. Its hard to take yellow offf a white canvas.
Sound is easy to fix if you know what you're taking away. So, with your EQ, start by lowering the level altogether to about 1/3 of the volume. This takes away some of the bottom, or bass levels. Then start raising the levels of what you do want. This works well if you are using a 20 or 27 band(1/3 octave) EQs. Those heavy piano notes are around the 300 to 415 Hz. Lower those first. The female vocalist is around the 2.5 Khz. raise those levels. Its all about what sounds good on the speaker system you're using. Many Recording engineers use 3 or 4 sets of speaker systems (one at a time) to get an extreme limit and compromise limit. After the project is mixed they might even burn it, take outside to their cars stereos, and get another source of sound to compare, and then mix it again. Experiment!

Well, the real solution is to re-record the track. It's heavily compressed with a firm peak of -1db. While the current file isn't clipping, the the original was probably very mashed. There's no good way to fix this. The goal here isn't to make the track good, but rather to make it "less bad."

Best: re-record the track. I'd re-record it even if it weren't mashed because, as lovely as the track is, the vocalist is a bit sharp at times and it makes the corner of my eye twitch. ;-)

Option 2: Use a parametric EQ that lets you zoom in on the problem. Ultimately you're trying to enhance the underlying track and you just have to live with the mashed up crunch of the original recording. It's there. There's no getting past it. Just ignore it as much as you can.

A bad recording is a bad recording and you have to either live with it or make a concerted effort to rerecord.

If it's going to take 2 hours to re-eq, balance and fix a track to make it sound only slightly less bad, and 3 minutes to re-record it, which would you rather do?

Subject: Audio Challenge
Posted by: Jim H
Date: 3/4/2006 7:43:10 AM

I have a recording from a video camera that was too loud and is clipped throughout. I'm looking for a way to make it more tolerable.

Rerecording is not an option.

Looking at this on the averaging function of DC6's spectrum analyzer, it appears as though the real culprit is the bass part of the piano keyboard. The frequencies below a couple of hundred Hz change very little in level and the harmonic distortion measures between 30% and 90%. Not a good thing. I tried to EQ on the paragraphic equalizer and lower the offending levels, then run the declipper. It helped some, but once you get the levels adjusted, you're left with the IM distortion in the voice. Not very encouraging.

Jim,

I can't seem to open your sample for some reason - and i wanted to download it and try something.

Do you have Sound Forge and the Noise Reduction plugin?

I recently made a similar (from your description) file a lot more listenable in Sound Forge by using a combination of the Vinyl Restoration and the Pop and Crackle filter. I played with the Crackle filter as the first filter in the chain to reduce some of the distortion crud, and smoothed it out a bit using the Vinyl Restoration filter. I then used a gentle EQ to boost the high end slightly to compensate for the loss caused by the filters. It did not restore the file, which was a recording of a vocalist at church who overloaded the input to the camcorder on an unexpectedly loud part of the piece (as well as the piano being REALLY load at the time), to pristine quality, but I was amazed at how well it did turn out. I thought I was stuck with it. Looking at the waveform, the tops of peaks during that period were perfectly flat. Somehow, the crackle filter managed to deal with it in a manner that worked, as it sounded to my ear as fuzzy crackle. Before using the filters, I lowered the clip volume about 6 db to give the filters some headroom to work with.

Like one of the other posts said about yours, this one had a lot of IM distortion, particularly in the lower midrange from the piano, and the uper mids from the female vocalist, so it wasn't just pure harmonic distortion to work with.

I wish I could download your file and try that technique on it.

Tom

EDIT: Well that time it let me get to it. Let me play with it and get back to you.


T

Jim,

It's my considered opinion that you are hosed on this one. It is far more distorted in spots than the one I had.

I tried the method described and only made slight improvements. I also tried another direct-x crackle plugin I had which worked better, but the improvement is only marginal. On the lightly distorted parts it did clean it up considerably, but it did nothing substantial to that hugely distorted portion mid-way through. On mine, most of the overload distortion was in the upper mids from the female vocalist and was easier to deal with. Yours had the most problems with the piano, and the IM and harmonic distortion would be virtually impossible to get out, as it's intertwinned with everything else.

There might be a way that you could split the file into multiple frequency bands, work on each seperately and sum them back together, but I don't think that you will ever get rid of that overload sound enough to make the result worth the hours of effort.

Hope you can find a solution.

Tom

just trying to do something nice for the artist.
==========
I tried for hours and can't do anything with it that doesn't sound terrible. If she is looking to use this for anything it will work against her.

Send the artist a tape or CD of the track and have her do the vocal again acapella into a tape recorder (even a portable one) while listening to the track you sent her through a set of headphones.

Have her record it a few times so that when she plays them both back the vocals match exactly in terms of synch. Then have her send you the solo vocal track and you will be able to lower the original and mix in the new vocal track. It will sound 100% better.

John

This was a spontaneous live performance memorialized on video. Any improvements are better than none. The EQ I've done so far sounds better, but like everyone has discovered, you can't get water from a rock. You should see the video [gasp].

what you have here is a variant of intermodulation distortion. The excessive bass in sections exacerbates the problem (psychoacoustically, anyway). I do restoration as a living, and I use a $70,000 workstation to do it, but there is no known cure for intermod. Do what you can using EQ on the low end.

Reid C