Help with Mixing.Mastering Top Singer Concert, piano & vocal

[NYCNewbie]

I have been passed along an informal tape of a concert performed by a well-known singer (by some at least -- if you know her work, you'll know immediately who it is). Without going into the backstage details, she's aware that the tape was made (not by me, but I've volunteered to try to do some work on it), and that someone is working on it. Again, without going into all of the back story, it's farily likely that she won't record again -- I could be surprised but it took her 19 years between studio recorded albums.....

It's vocal, female obviously, with piano only. Cabaret setting - 186 seats (if you want to see the entire audio setup, you can look at this link: http://www.therrazzroom.com/Venue.html.

I have included some clips you can find at
http://sites.google.com/site/homerecordingmixhelp/
then look for the one page link on the side that says Downloads for Home Recording Readers
(i couldn't figure out a way to post audio clips on this site, sorry)

This concert was taped with a tiny little recorder so I'm told, but all in all, it could be much worse. The clips include a snippet of Bacharach's Anyone Who Has a Heart, and two snippets from a Bacharach medley which includes Alfie -- the first clips has fire engines going by, and the second has coughing.

Over all, the tape like i said is surprisingly good -- it seems a bit bright/harsh, you've got your usual things to contend with (fire engines, chairs moving, very loud applause, people coughing). And with the recording device, I have one track with which to work.

I'm a great self-teacher, and I read through a lot of the posts on here, and I read through an Adobe Audition book which was okay but without any real 'advice" in it. If anyone would be so kind to share an appropriate workflow for a recording such as this, suggest any appropriate (free) plugins for Audition, and offer any thoughts on e.g., how to get rid of sirens, and bad noise or eliminate it as much as possible.

It might be helpful to know that when she was recording, Michael DeLugg engineered much of her work (Delugg, aside from this vocalist, is best known for engineering a good majority of the Barry Manilow albums -- despite what you think about Barry (and I think most secretly like him), Delugg gave all of those albums a particular sound which worked so well. Similarly, his work gave the vocalist on these clips some real space around her voice, and there are a good handful of her songs and their work which I still consider some of the finest pop songs ever recorded. On the other hand, producers like Michael Masser suffocated her in wildly orchestrated songs, and with him and others there just wasn't the life in her singing that her oft-time producer and Delugg did. The album 20 years later was produced by Steve Addabbo (Suzanne Vega fame), and while much different that the delugg sound, gorgeously produced.

All of this is mainly to see if a CD could be pulled out of this recording, flaws and all, that is, a CD for the artist to sell or MP3 download to generate some # as she made milllions for Columbia but never made a dime of royalties.

On a separate subject, I also included in the link above a snippet of someone who has posted on youtube his remastering a number of the quadraphonic releases of songs by the Carpenters, and they are wonderful. You can find t them all on youtube, and he uses Audition for his work. Any ideas about his process in making these quad mixes sound so terrific?

Apologies for this being so long, but i thought more detail than less would be better. (P.S. I have nothing to gain by any of th is, it's all just pro bono effort. My Audition edition is 3.0.

Thanks very much.

 

[oretez]

If your source is a single track containing voice, piano plus unwanted noise there is no 'mixing' in the traditional sense. My first suggestion is to drop in on this forum that is focused on AA and search for 'restoration' threads paying some particular attention to posts by Andrew Rose. An archive search of old Syntrillium (Cool Edit) threads discussing the 'Younglove' approach to noise reduction might be helpful as well.

A wrinkle added in AA 2 (I think) was frequency space editing which can be of some help with isolated unwanted artifacts (e.g. coughs) generally speaking the siren, representing a significant range of frequency and dynamics, is going to be a bit more difficult to even reduce impact slightly. Artifacts with attack transients comprised of limited frequency range with attack/decay that is short and sharp are much easier to mitigate.

As far as achieving results that are marketable, that really depends on a couple variables about which I have no opinion (or information concerning). 1. mp3's provide little accurate information concerning the frequency and dynamic range of the source. 2. artist client expectations 3. how rabid the fans are

(what is considered one of the best live jazz recordings (to date) has many of the same deficiencies as your source appears to have. It sold reasonably well in the early 60's and continues to draw at least a handful of new fans today)

If the source was recorded in any lossy compressed format (MP3, etc.) I wouldn't approach this as anything but a learning exercise. If source is relatively decent (e.g. 44.1/16 or 24 wav) general approach is to first make sure originals are safety stored and backed up. Then typically I'll start by breaking 'show' into smaller units . . . typically song by song. This is done for a number of reasons but partially because most accurate edits are, very generally speaking, executed on the smallest chunks of data (it is pretty easy to reach a data chunk which is simply so small as to exceed any cost/benefit . . . so, as with all audio editing, everything is compromise) Another reason to break a 'show' into 'tunes' is that it is pretty easy for any edits to alter what might be called the audio ambient. Assuming that individual tunes can have different emotional/performance profiles breaking a show into tunes can be fairly decent bracket for processing.

I process in floating point so somewhere in this initial triage I'll convert whatever the source is to 32 bit floating point wav file. Depending on the project this might be done to the large working source file or executed as a batch on individual tune files. Next I'll look/listen for artifacts that can be addressed across the entire file. Clicks, 60 cycle hum, back ground noise that is at a low enough level to be considered to be consistent across a data chunk. Most DAW's have 'tools' (i.e. scripted presets) to deal with a number of this type of artifact. Removing artifacts almost always has chance to introduce new ones. Even if you locate a segment of background noise, for the noise profile, that is free of isolated percussive 'plosives' and seems to be consistent over the course of defined 'data chunk' too aggressive a setting for removal will introduce new artifacts. And generally speaking in most 'live' recordings pristine studio sound should not be mandatory . . . as a rule some background noise is preferable to new digital artifacts. The Noise reduction algorithms of different vendors tend to be slightly different. I find AA's to be quite good and if I only had a choice of only one, AA would probably be the one I chose. As I don't have only one it is not unusual for me to use both AA noise reduction and Izotope's RX suite (whether it is simply personal preference or not I do prefer Izotope's dithering algorithm) The two apps seem to work in slightly different ways, clean up slightly different things. Additionally, depending on specifics of the source I might run more then a single pass AA NR. One with a relatively small FFT size, 4096 for example, one with a large FFT size (as large as a cost/benefit for processing time will permit . . . with modern machines this can easily be maximum) FFT size determines the size of the 'window' or data chunk being processed, smaller FFT numbers use bigger windows. The other six parameters, in AA, are simply things you have to experiment with. While some can be set consistently across more then one project others will be specific to a single tune. And most are subject personal preference. I seldom find it beneficial to set noise reduction level above 50%.

After noise reduction I tend to normalize volume to 0dB. I do this primarily so that I start work at a consistent volume 'profile'. There are individuals who do more restoration work that probably do not take this step. And there are those who execute NR after normalizing. And there are reasons for executing NR after, but the choice can be dependent on specific source. After NR and 0dB I look for artifacts that can be removed manually, coughs etc. Stuff that can be identified in frequency space editing and processed with 'healing' brush, gain reduction, EQ, dynamics processing, etc. Then there are some things that can possibly be addressed via old school NR, companding. Mic bleed, for example while really not a candidate for profile NR might be amenable to a little compression/expansion treatment. Even something like a siren might be susceptible to this approach.

All this tries to deal with stuff that an audience can identify as obviously not belonging and might find distracting. Unfortunately this gets to the point of addressing what I see as a far more difficult problem with your source, imbalance between piano and vocals, with piano tending to mask vocals. and there are no simple ways of dealing with this. On a strictly commercial project it is probably something I'd try to avoid working on. It is basically a thankless job that except in very exceptional circumstances does nothing to enhance a reputation and only in even more exceptional circumstances are clients willing to pay appropriately for the effort. But there is stuff you can do that tends to fall into the realm of re-synthesis rather then merely editing. Just this side of re-synthesis, with material where attack transients of piano and vocals have distinct dynamics and frequency range, source can be broken up into overlapping (sharp delineation between bands will introduce 'ringing' artifacts, for example) frequency bands. It is seldom necessary to define a full range of 1/3 octave bands. The bands typically are copied, one for piano one for vocal and then processed via compression, expansion, EQ and gain then eventually re-mixed, reassembled, frequently with foldback from the unprocessed track. Sometimes it works. Sometimes it just creates an endlessly evolving mess.

good luck.

 

[NYCNewbie]

thanks so much for your reply......i will read it in detail and look for the other helpful posts.....as for the fan base, they are absolutely rabid (as in this artist could sell out Boston's Symphony Hall and other similar venues, with virtually no product for over a decade, she is that much beloved by so many.....)

if anyone else wants to pitch in with ideas, they'll always be appreciated.'

thanks so much

 

[NYCNewbie]

and yes, the CD came to me via MP3 form, which I've converted up to wav and have separated each song, monologue into separate wavs on which to work -- because the likelihood of the singer pulling together a recording deal (with all of its budget requirements, artists' charges, trying to recover whatever advance may have come her way, and the whole get go) coupled with an artist whose moods and opinions tend to fluctuate wildly, i don't think that will happen.....she had another live album and DVD put out some six or so years ago, and while it was fine, this admittedly uneven low-end recording shows a lot more life and spontaneity than the earlier ones......there's a built-in market and fan base that simply refuse to disappear for this artist, and rightly so, so the perfection of some of her studio albums for Columbia isn't expected or anticipated, but virtually anything would be better than nothing, both for the artist and the rabid fans......I mostly wanted to reach out to you experts before I launch in -- smart, a fast learner, etc.-- who might immediately have some good thoughts to help me.....

thanks again so much!

 

[oretez]

if it is at all possible to get a not compressed (lossy data compression, i.e. MP3) version, prior to spending a lot of time on the project I would suggest making the effort

MP3 encoding changes the information you have to work with drastically (and even while lame variable bit @ 320 bps can sound indistinguishable from redbook CD (in some environments) you still lose info that is not recoverable (and re-synthesis while it might improve listening does nothing to preserve original performance) and you need that lost info for the kinds of edits necessary to improve recording appreciably

Additionally since you are talking about distribution via down load re-encoding the performance back to lossy format (i.e. MP3) simply compounds the problem

but again 'success' basically depends on what fans will accept and there is a lot of really bad audio distributed successful on net so 'standards' will inevitably change