"Fixing" a choral piece

I recorded a choir about a month ago - it went quite well and the recording turned out pretty well.  The choir director recently had a chance to listen to it and said there are a couple spots he'd like to "fix" where the sopranos are not singing together and are not on pitch.

My first thought - and what I told him - is that we should just get the singers back in the room and punch-in over the sour spots.

I can see trying to notch out the different parts and maybe applying fixes to narrow frequency bands, but my heart tells me this would be time consuming and probably futile.

Is punching in the only real option here, or are there tricks I might try to "fix" a choral recording?

I'm assuming these measures aren't repeated exactly elsewhere in the piece for a cut n' paste?

I don't think you'll be able to re-enact the gig close enough to make a freshly recorded punch in transparent, but maybe that's just because I doubt *I* could do it...  

I always try to get a couple takes of difficult segments, but of course, sketchy performances can even happen on the simplest, best known passages.

It wasn't a live show - it was a recording in a rehearsal room, so conditions can be duplicated quite accurately.  There would likely be a few different voices, but the choir is large enough I suspect it could work if I'm careful.

I have several takes of each movement, but I think the bits he's picking on aren't much better in the other takes.  Also, there was a flute playing in the room.  I close mic'd the flute but there is lots of bleed in the vocal mics and flute bit improvised some and did not play the same with each take - I don't think a cut and paste would work because the flute would double up.  If we setup again and I punch in vocals only (the flute was from out of state and getting him back is not an option) I can probably add some extra reflections to the flute during the fresh vocal takes and squeak by.

Karl, I've had occasion to have to do this twice, once with a single word, the other with a 5 second vocal phrase.  I chose not to punch in.  I found doing several re-takes, each on different empty tracks, allowed me to use the volume envelopes to precisely cut the errant parts on the original track and at the same time bring in the best re-take. Highlighting and zooming in can certainly get you to where you are on the money.  Recording the re-takes on separate tracks rather than just punching in over the original gives, at least for me, more flexibility when replacing the bad with the good.

Another plus for going on different tracks: if, for some reason, the acoustics or eq on the re-takes are not quite the same as the original, then you can treat the re-takes however you need to, to get a match.

Cal got in just before me, but to reinforce what he said, I would NOT punch in...I'd record the new material in a seperate track that I can play around with to match the original stuff as nearly as possble, then mix between the two.  I think you could make it pretty seamless this way if you're careful.

(FYI I recently did similar, albeit with a solo vocalist, who came in and fixed a couple of notes almost a year after the original recording.  It worked!)

Bob

I would agree that this advice is right on the money.  

Karl, I know you're aware of much of what I'm going to write here.  I'm writing in more detail than you need in case others want to give this a try.

If the tempo is slow enough, you can delete 20ms or so of the leading edge of something that's not together (auto crossfades enabled, of course), but there's not much else you can do to it.  If the note begins with a sibilant consonant, the "Fill Single Click Now" can do wonders with one or two who are early, though miracles can't be counted on.  

If some sopranos are correct in pitch and others aren't, you can do an extremely narrow notch with the parametric EQ to reduce the flatness.   Make sure that Options/Settings/Data - Smooth Edit Boundaries is enabled to something fairly smooth like 50ms.  (Set the Parametric's Low- and High-pass to neutral.)  Enable one band, set the Q to around 1 or 2 Hz, then "ring out" the flatness.  After isolating the problem, setting that band to around -15 to -20dB will kill it quite viciously.  Each overtone needs to be a tiny bit wider in Q and slightly less cut.  Multiplying the first frequency for the other bands, but ring out at least the first overtone because the exact double may not always work.  Also beware of needed overtones of other pitches that intersect with what you may need to filter.  Few things still sound flat, particularly sopranos, if five harmonics are cared for.  This technique is also helpful for first violins playing around the 39th position of the E string and one of them "misses" the mark a bit.

Time consuming? Yes.  Much better to get a useable retake.  But I have a whole slew of presets for this kind of filtering, and after awhile it actually saves time -- so long as the sopranos aren't terrible.  It really comes down to how much time you think either of these approaches might take.

I chose not to punch in.  I found doing several re-takes, each on different empty tracks, allowed me to use the volume envelopes to precisely cut the errant parts on the original track and at the same time bring in the best re-take

That was the approach I was planning to take.  I'm thinking that if we can incorporate the fixing into a session where we record some additional material it would it would be pretty painless.

I'll see what I get back from the choir director before I spend much time on it.

 Each overtone needs to be a tiny bit wider in Q and slightly less cut.  Multiplying the first frequency for the other bands, but ring out at least the first overtone because the exact double may not always work.  Also beware of needed overtones of other pitches that intersect with what you may need to filter.

And are these frequencies found through the Frequency Analysis window, or by ear (and what do you then hear?), or by math after finding the first freq?  This amount of treatment is something I haven't done, haven't had to... If you use the ParaEQ to ring out on, say, the first harmonic, what will I hear... an octave, a fourth, or....?  How do I know I have the flatted overtone and not another?

Ah, it's the ol' ear thing.  Hearing intonation pretty well is essential to this parametric EQ technique!

When I say "ringing out," the boost is huge, as much as 30 dB depending on the material, so when you "hit" the note there's no missing it.  Plus some of the math is easy: soprano high A is 880, first space E is 660, middle A is 440 etc.  So the hunting and pecking is partly knowing what pitch is being sung (via score or hunting it down on a piano), partly a calculation of the frequency, and the rest is hearing it.  

Once you've found the note, nudging the pitch arrows up and down during preview can be quite illuminating.  You'll hear how some voices are on, some are flat, etc., if the Q is super-narrow.

I have the score - or my Daughter does - she's one of the singers.  I messed with trying to notch out the wrong notes and was not impressed with the results.  I may mess with it a little more tonight.

Zemlin, I suspect when you tell the choir director what's required to fix the problems, he'll say "ah - in that case, it's fine, we'll leave it as it is" - least, that's often what happens in my experience.

Karl, the trick is to make it as un-invasive as possible.  

No preset is going to fit a given need exactly, but to furnish an example, here's one of my typical presets.  Copy this to an .ini file on your desktop with Notepad, and drag it onto an opened instance of CEP or AA and it will add it to your presets automatically.

[Parametric Equalizer]
Item57=C high,3,0,40,0,18000,0,1,1,1,1,1,1046.3,2105.2,3142.8,4212.1,5260,1,2,3,4,5,-32.7,-28.4,-27.6,-21.1,-13.8,0

I know i dont have much experience in this specifically, but I don't punch in. What I do is go down one empty track, record earlier, and after the point of problem. Then i use volume envelopes until the change is almost seamless. This procedure has worked almost all the time in any situation involving vocals... Just my idea.

RonC

Since I don't record in AA, the logistics of punching in are different - the details are different - the end result is the same as y'all are talking about.

really? what do you use? Ive always thought you used cep/aa?

RonC