In Part 1, we asked: Audience or no audience? For me personally, the answer is simple:I must have an audience to play my best.
For the rest, I do as suggested to me by pianist-producer Lincoln Mayorga:
One: Record a concert twice in the same hall, a day or two apart.
Two: From the recordings, choose the better performance of each work.
Three: Identify any fatal errors in the chosen performances.
Four: Correct the errors if possible. For instance, if I goofed in a certain passage in one concert, maybe I played it fine in the other concert, and can “cut and paste.” Or maybe the same passage appears again elsewhere in the piece.
Five: If correction to a level I consider “releasable” is not possible, I conclude that I wasn’t ready to record that work! (But things often work out OK.)
My first three albums indicate a range of what can happen. The first (Scarlatti/Beethoven; out of print) came from a single performance with no editing. I played the concert once, and lucked out: it was releasable.
The second album (Schumann/Chopin; out of print) had one edit, to eliminate a cough between two chords at the end of “Kinderszenen.”
The third album (Prokofiev 6th Sonata; LP out of print; CD available) came to grief due to a fault in the microphone, an experimental model. I recorded again with a properly-functioning mike and an invited audience, in a process halfway between concert and studio recording. Take 1 was a straight-through performance of the entire piece, all four movements. Take 2 was the first movement again, by itself. After listening to both, I decided to use the 1st movement from Take 2, and the rest from Take 1. A few more takes, very short, were “covers” for small errors; and we were done. The audience listened with involvement to all the playing, and waited patiently while I listened to all the takes. The experience made me realize, though, what a heavy demand even those few takes put on them.
Later albums I recorded in a pure concert situation as described above. For “Pictures at an Exhibition” (LP, CD), I edited with Hugh Davies, who for 40 years was Angel’s top editor. The performance is 31’25” long, of which 30′ was taken from one concert, and a few bits totaling 1’25” from the other. As we finished, I remarked that this was by far the most editing I’d ever done on one album; and Mr. Davies exclaimed, “It’s one-fourth of the least I’ve ever done!”
In other words, when we are ready to perform, we’re ready to record, and are then rewarded by needing little or no editing. (Nor do I recall any reviewer complaining of more editing being needed.)
Can recording this way work for ensembles? It seems obvious that a quartet, an octet, a full orchestra must be more likely to produce errors than one player, other things being equal. But, as one great mathematician liked to ask, “Is it so obvious?” Or can an ensemble indeed play well enough for release, whether because they’re individually just that good, or from the inspiration of playing together, or the excitement of recording? Or because occasional errors are masked by thicker textures?
One answer comes from Royal Philharmonic’s live concert recordings: apparently it can be done! Another, and definitive, answer, comes the existence of direct-disc ensemble recordings. Like the 78s of long ago, direct-disc LPs were cut in complete sides, with no editing possible except the choosing among multiple takes of complete sides. Sheffield Lab released three by the Los Angeles Philharmonic with Erich Leinsdorf, three by the Harry James Big Band, one by the Chicago Symphony Winds (an octet), and albums by pop, jazz and bluegrass ensembles, and the Kodo Drummers of Japan.
In direct-disc recording, each take is a whole record side; and each side has to be good enough, as in concert recording; but as in studio recording, you can do it again if need be. The unique additional element is performers electrified by the pressure of the situation; a substitute for the excitement of having an audience.
LPs are obsolete, but what we’ve learned from these experiences still applies. For instance, in one recording by a blues ensemble, the recording level was set and then everyone took a break. When they returned to record for real, their level was 15 dB higher than before! Subjectively, that’s more than twice as loud. Electrified indeed! So perhaps it is plausible that an ensemble can do a successful concert recording. 
