Menu
What's new
By:
Filters

The Adventures of Creating a Choir Library - January 8, 2024 Update - Why, Oh Why Did We Do Legatos???

OUR STORY CONTINUES!

Back when I wrote my April update, it was embarrassing enough to concede I wouldn’t be finished by NAMM, but I at least figured we’d be done before summer, along with a triumphant new entry in this thread. Well, that didn’t happen, so let’s make that end of summer. Oh, did I say end of summer? Let’s make that before Black Friday. For sure!

Ah, those days when I was young and believed silly things. Now that I’m older and wiser, I’ve set a more realistic goal of having this ready by NAMM. No way I’ll miss that deadline! Wait a minute, NAMM is just a couple weeks from now? ~sigh~

SO WHAT’S BEEN THE HOLDUP?

Before I get to the stuff you actually came to thread for (the “how to build a library” part of our story), “life” has a way of complicating things. Nothing bad, mind you, but after decades of chasing career goals, and prioritizing work over other stuff (although can making music and having fun creating libraries really be considered “work”?), I came to realize I need to lead a more balanced life. Or more accurately, my wife instructed me that I need to lead a more balanced life.

So we’ve been doing a lot of traveling, theater, events … if she’s bookin’, I’m goin’. And she’s right, it’s nice. (Two and a half weeks in Ireland was great!) Just between us, though, can I admit that when we’re listening to musicians in a pub, or at Disney Hall, I sometimes start thinking about what articulations would be needed to recreate what we’re hearing? Yeah, you know what I’m talking about, right?

Another factor is we started a studio renovation this summer. Back in 1993, I scrounged up every nickel I had, then borrowed a whole bunch more nickels, and built this place. (By “built,” I should clarify that this was a 3,000 square foot warehouse I leased and I built out the interior. We didn't buy the building until later.) It was a big risk, but I was producing records at the time, and realized I could double-bill the record companies by collecting a producer fee plus charge for the studio, too. Greed, thy name is Mike Greene! Plus I could hire engineers and rent out the studio for extra income. (Which I hated, and as soon as loans were paid off, I stopped that.)

Anyway, certain choices in decor I made back then haven’t aged well. Plus flooring isn’t supposed to last 30 years, so I consented to hiring a decorator and we’ve been re-doing the whole place. New paint and flooring throughout, cabinets, lighting, artwork … might as well do it all. That way when friends come over, I won’t have to keep making excuses for, “Here’s what happens when you put an adolescent straight male in charge of decorating.” (I don’t care what my wife says, I thought it all looked great at the time!)

What a massive undertaking this is. For starters, there was way so much junk here that creeps up on you. Things that need saving for a few years are a little less critical after a couple decades. (I sure hope nobody calls to ask if I still have their old 2” tapes!) I have to personally stay involved in all that, because otherwise I can only imagine hearing, “Hey Mike, you had too many mics, so we tossed the old ones.”

It’s been a few months, floors and paint are all done, and now we’re on fixtures, furniture, and artwork, but I had to put it on pause, since it was stressing me out too much. Especially as it drove me insane that I was spending so little time on the choir project. I do have talented people working with me, of course, but this was a textbook case of the negative side of “leading by example.” In our weekly Zoom meeting, I’d too often say “I didn’t get a chance to work on anything this week,” and that became contagious. Well, maybe not "contagious" so much as frustrating when "Mike doesn't seem to giving his all on this, so why should we?"

So that’s part of the delay. (Although reading it back, it might be a little overstated. Those things took a lot of time, but not all my time. I have been working.) The biggest part of our delay story is …

WHY, OH WHY DID WE DECIDE TO INCLUDE LEGATO???

We recorded legatos, but there was still a question as to whether to actually include them. Now, I know what you’re thinking: “But Mike, that’s crazy! If you already recorded it, why wouldn’t you include it?”

Two reasons. First is that until we actually test it, we don’t know if it will actually sound good. You don’t want to release a library with weak legato, since many people will judge a library entirely on that. And many of those people will make YouTube videos, or post on forums, tossing around words like “unusable.” (You know what I’m talking about, right? ;) )

So until you know a major feature like legato will actually sound good, it’s best not to promise anything. Too risky. Heck, there’s no legato at all in Sunset Strings and it’s selling just fine, so it's not like it's a must-have for the choir.

The second concern is a bigger one. That is that the legato, at least as I’ve designed it for this library, is a massive editing task. I mean massive. Not just in the chopping up of samples, but also in the tone and level matching. I want this legato to be really, really good, so there’s a lot that goes into that.

Rewinding a bit, I’ve done legato before with Realivox. That worked out well, I believe mostly because I spent a lot of time volume balancing the samples. A volume (or tonal) mismatch in a legato transition sticks out like a sore thumb, and IMO, most libraries with reputations of weak legato could improve themselves by just going back and doing some level matching. Easier said than done, though, because that means going through each sample, one at a time, listening to how its volume compares to its starting and landing counterpart samples. It’s very time consuming.

It’s also not something just any sample editor can do, so this part of the editing process is delegated to the A-Team. Which is also the more expensive team. Truth be told, I can’t resist doing some of the editing myself, especially in the beginning stages, where I want to hone in on what the best process would be. Which is ...

continues ....
 
SO … WHAT EXACTLY IS THIS PROCESS?

I’ve been debating with myself whether to share this part, since there’s a bit of secret sauce to it. But then I started thinking the real “secret sauce” isn’t so much the “how,” but rather it’s the “time,” and the fact that I’m willing to actually do it. (Or rather that I’m willing to pay people to do it. At A-Team rates.)

In fact, there have been times I’ve shown other developers how I put together Realivox Blue and they’ve said, “I’d never have the patience for that.” That’s not a knock on them that they choose other ways to spend their time, mind you. This is the sample library business, not the sample library lets-hope-our-ASCAP-checks-are-enough-to-cover-our-losses-because-we-spent-so-damn-much-time-on-this business. Honestly, from a pure dollars and cents business perspective, I’m not so sure this legato thing was such a good idea. We’ll see. (Spoiler alert, it’s sounding great, so in answer to my earlier concern, we will indeed include it. But will it result in selling more copies than what the legato cost? I don’t know.)

Anyway, with Realivox (both Blue and the Ladies), the recorded legato intervals are just a 2nd up, a 4th up, a 2nd down, and a 4th down. (Surprise!) All the other intervals are covered with just those samples. (And some fancy scripting work. Sometimes I impress even myself!) And it worked! Or at least it's “good enough,” since I have yet to hear anyone complain that the legato samples don’t sound genuine across all the chromatic intervals.

Two things, though. Realivox is solo voice. And it’s dry solo voice. You can’t use these scripting tricks with an ensemble (for starters, they don’t all slide at exactly the same time), and especially not an ensemble recorded in a large space, since telltale room ambience from the initial note bleeds into the transition.

So for Sunset Choir, we had to go chromatic, an octave’s worth of samples in each direction. That’s a LOT of samples! Here’s the math:

1. 12 up intervals + 12 down intervals = 24 legato samples for each note.
2. For the sustain samples, most choir libraries are recorded in whole steps, so we decided to do the same. (Realivox Blue/Ladies/Men are chromatic, but if the big boys are doing whole steps, and if recording time is costing me thousands per hour, then by golly, whole steps it is!) So that’s 15 “landing notes” to fill the range. So now we’re at 24 x 15 = 360.
3. Men and Women recorded separately. (360 x 2 = 720)
4. Three speeds: regular legato, slow bend legato, and slower bend legato. (720 x 3 = 2160)
5. Two dynamics (mp and mf) for the regular legatos: (2160 + 720 = 2,880)
6. Three vowels for all of them: Oo, Ah, and Mm. (2,880 x 3 = 8,640)

I don’t know about you, but where I come from, 8,640 is a lot of samples. All of which need to be edited separately, sample by sample. And when I say “edited,” the chopping part is just step 1. Here’s a rough explanation of the full process:

1. We recorded our legatos in a similar method as what I think Jasper does (although I’m not precisely sure about his process), where I had them sing these various melodies I wrote. Four of these “melodies,” which, when you deconstruct them, include all the intervals we need. (It’s like that typing exercise: The quick brown fox jumped over the lazy dog.)

2. Editor #1 chops those performances into short snippets of all the legato intervals, then sorts them into folders for each landing/destination note. (The “landing note” is the note a legato interval lands on.) Inside that folder, the snippets are sorted lowest to highest.

3. Each vowel/sex/dynamic/speed combination gets its own Pro Tools session. There are 24 tracks. (72 tracks in reality, since there are three mic positions, but let’s call it 24 so it’s easier to understand.) The first track is “Up 12” (12 half steps = full octave) the second track is “Up 11” (11 half steps = Major 7th) and so on, with the last track being “Down 12.” (Actually, there are more than that, since some intervals got duplicated across those four melodies I wrote. So for example, we have two “Down 12” tracks to choose from. In reality, we have 29 tracks.)

The landing pitches for all the snippets are laid out with lowest note at 0:00 on the Pro Tools timeline, next note at 0:20, then 0:40, then going up the scale to the highest note as we keep moving up the session timeline. So if you think of this PT session as a grid, the landing note pitches go left to right, and the intervals go bottom to top.

Now that we have a Pro Tools session filled with all these legato snippets, we then add the sustain samples, so the legato snippets can be stitched (crossfaded) into the sustain samples. Editor 1 then raises or lowers the volume level of each legato snippet so it matches the sustain sample it will be blended into.

4. Editor #2 (this is higher level work) will crossfade the legato snippets into the sustain samples. First he adjusts the volume of the legato snippet so that the landing section matches the volume of the sustain sample. (Editor 1 got it close, which saves preliminary time for Editor 2. We want the volume-matching end result to be as close as possible.)

4b. He may also need to adjust tone/timbre of the legato, if it's not a perfect match to the sustain. The legatos were obviously recorded separately from the sustains, and I tried to get them to match timbres from one recording to the next, but with vocals, “Sing it exactly like you did yesterday” is never going to be perfect. Especially when they start singing these legato melodies, where with those brilliant masterpieces I wrote, how could they not get carried away with emotion!

4c. I should mention that earlier in the process (between steps 1 and 2), we tuned all the landing notes in those melodies. Otherwise there will be a audible pitch jump when it crossfades into the sustain. (You don’t want to merge a sample that’s 20 cents flat into a sample that’s at 0 cents.) That tuning process is ultra-tricky. (I delayed sharing that step until now since it would have been harder to understand earlier.)

5. Next, he’ll adjust the transition points and find whatever crossfade sounds smoothest. This is by ear, and it’s not a fast process.

6. We use “Consolidate Region” to create new samples which combine the legato snippet with the sustain sample. With some libraries, the legato samples are separate from the sustain samples they go into, but I wanted to eliminate all potential for mismatches (volume jumps or whatever), so I do the combining here, so the legato samples includes the sustains. Note that the sustain volumes with these legato samples are identical to the sustain volumes in the regular sustain groups, so everything in the final instrument should play smoothly.

7. Map the samples into Kontakt and marvel at what a great job we did.

That’s a lot of steps, right? But … uh-oh, we ain’t done yet.

8. We discover a few surprises when everything is mapped into Kontakt. The main one being … well, this one is hard to explain, so consider this example:
* We have a women F4 sustain.
* We have a women F3 sustain.
* We want to play legato from F4 to F3, so we hear the F4 sustain sample, then we hear a legato sample that goes down from F4 to F3.
*The F3 sustain part of the legato sample perfectly matches what a standalone F3 sustain sample would sound like. (Which it should, since they’re the same recording. Remember, we stitched the F4-F3 legato onto the F3 sustain in Pro Tools.)
* But the F4 part of the legato sample is louder than the sustain F4 sample. Uh oh! That means if you play an F4 into an F3, the legato has a big volume bump at the start of the transition. Yikes!

Truth be told, we kinda expected this. A dirty little secret of choir libraries is that there’s usually some volume balancing going on across the scale, since, especially with women, they sing their high notes much louder than the low notes. (Good luck getting sopranos to sing a high A at mp!) So for the sustains, we lower the volume of the high note samples, since it makes for a much more playable instrument. Purists may be offended, but unlike a live choir, which has ways to compensate for imbalances like this, we’re dealing with a sample library, which is less flexible. So the bottom line is do you want “pure” or do you want “sounds good”? (Rhetorical question. You’ll get what we give you, dammit!)

So we need to tweak (volume curve) those particular legato samples so that the transitions sound smooth. It’s an annoying extra step, but fortunately this is only an issue for intervals close to an octave. (Intervals up to a fifth or so are fine.) So that takes us to our final step:

9. Load those tweaked legato samples into Kontakt and now marvel at what geniuses we are!

Great! You've completed one set! Now repeat that process for all the other vowels, dynamics, men/women and speed combinations. Don't you wish you could be an editor and have all that fun?

... continues ...
 
BUT MIKE! COULDN’T YOU RECORD LEGATOS WITH SUSTAINS?

Yep, we could. In fact, that’s how we recorded legatos with the Realivox Ladies and Blue.

There are some things to consider with this choir project, though. Using the “melody” method, the legato recordings went fairly quickly, since a whole bunch of intervals got recorded in a single pass. I say “fairly” quickly, though, because these still weren’t fast. (Especially the slow bend legatos, which are really hard to get in sync. The success rate on each attempt of those was probably under 50%. That's a lot of retakes.) I’d guess we spent 15 hours total on recording legatos with “the melody method.” At a little under two grand an hour, call it $25k to record legatos.

Now let’s suppose we instead decided to record each legato with a full sustain attached. There’s certainly a good argument for this, since there’s no splicing involved, so there won’t be any dodgy edits, plus the editors can eliminate a couple steps, saving time on the back end. But …

There are 8,640 of these legato samples. If we recorded them one at a time with a held sustain, that’s gonna take some time. Bear in mind that with the melody method, it’s easier to keep everybody's note transitions in sync, since singing melodies is natural. But with one-off legato slides, it’s less natural, so there are a lot of timing errors = retakes. Then we also need to make sure the sustain section sounds perfect as well, so add more retakes on top of that.

Doing a little napkin math, that would be about 40 hours. Maybe a little less, but I doubt it. So we’re looking at maybe $70k?

And a lot of very un-fun hours, because recording legatos is the worst. The singers loved doing the other stuff, because it was so unusual and there was a lot of variety. But legatos? Ugh. It's all boring precision work, so I would have to be a cheerleader to keep their spirits up for what would be *days* of it. 20 hours for the women and 20 hours for the men, where their sessions are 3 to 4 hours each, that’s an entire week of legato recording.

There’s an editing issue to consider as well. Sure, the editor can eliminate the splicing steps from his process, so that will indeed save time. And money. (Although editors make a teensy bit less per hour than the $2k/hour that the choir recoding costs, so we need to keep that savings in perspective.)

It's not all roses for the editor, though, because he now has a new task of 8,260 samples he needs to loop. You see, with Mike’s Magic Method, where we splice each legato onto the existing sustain samples (the existing already looped sustain samples), we set the PT sessions up in such a way that the loops carry over into the newly created legato samples. No new looping needed!

And then there's an additional benefit to my method, where our legato samples are 100% consistent with the sustains, since the landing note *is* the sustain. So there won’t be some notes that get darker or brighter as you play a melody.

Inconsistencies can be a audible giveaway, since light/dark variations stick out more when playing legato than when playing chords. Vocals are sooo much harder to keep their tones consistent than with instruments.

It’s one of the reasons I have to be a cheerleader throughout the recording sessions, since there’s an audible difference between a smiling singer and a frowning singer. (Because of the mouth shape.) Years of recording pop songs taught me this, and in those pop sessions, I can’t even count how many times I've said, “Okay, the notes were great, but this time smile as you sing it.” (Try that with your own songs. Even if it’s not necessarily a “happy” song, it just sounds better.)

THE REALITIES OF WHAT GOES INTO A LIBRARY

The legatos have dominated so much of our time, which is a bit unfortunate, since that’s not really what the library is supposed to be about. (Our template, Sunset Strings, had no legato at all.) That’s not to say legatos are being done at the expense of other articulations. It’s more a matter of them pushing the timeline back. So the solo-voice repetitions, for instance, still aren’t done. Those are a complicated story of their own, and as I’m diving back into them, we still have weeks, at least, just for the unfinished solo voice repetitions.

Time carries a cost, not just in the cash outlay of writing checks, but also in opportunity costs, where it’s very difficult to make a profit with an 18 month release schedule. (Sunset Strings and Nightfall were each 18 months. And those were after Hip Hop Creator, which took almost three years!) I haven’t yet compiled my income and expenses for 2023, but I suspect it’s close to break even. After a net loss in 2022. It’s hard to get rich that way. ;)

Ultimately, you always want to create as super-duper-amazing of a library as you can, but from a business perspective, and even from a satisfaction perspective (it’s not fun to go this long without releasing), when that comes at a cost of zero releases in 2023, it has me questioning what I’m doing. This is a lot of effort for a business that’s essentially subsidized by my other sources of income.

Then again, I’m optimistic about 2024, and with everything we have coming up (I swear, this will finally be the year!), I’m even thinking it could be a huge year financially. With a little luck, “huge” enough that 2022 and 2023 can be considered “investment” years. Which actually is what I consider them, and is why I’m okay with spending so much.

I’ll need more than Sunset Choir to make 2024 a big year, though. As fate would have it, when it started looking impossible for Sunset Choir to be released before Black Friday, I decided to pivot back into the Realivox Men, thinking it might be doable to release the Men before Black Friday. (You really, really, really want to have released something new in the year before Black Friday.) Turns out that wasn’t doable, either, but I did make major progress on it before throwing in the towel when I realized the Men couldn’t make the deadline, either. So the men only need a month, maybe two, whenever I can get back to them.

So 2024 should be the year. Especially since there’s some other stuff in the pipeline as well. Color me optimistic. Although … anyone familiar with Realitone knows I’ve been optimistic before …

AND THEN THERE’S NAMM

This will be the second NAMM where I’ll be showing the same incomplete library. That kills me. The booth is $10k, which is kind of a waste, although as I explained in an earlier post, my NAMM booth is more of an indulgence than a "business move" anyway, so screw it, I’ll just try to focus on having fun at the show. I’ll show what we have for the video guys and maybe a few friends, of course, but then I’m just gonna be in hangin’ out mode.

I had hoped to have an “In Progress” version at the show for the people who want to buy it now, similar to what we did with Sunset Strings and Nightfall, where you could buy the mostly finished unencoded version, then you’d get the completed KPlayer version (for free) when we released it a few months later. Essentially a pre-sale, with the added bonus of getting an “in progress” version early. That worked really well.

I can’t do that here, though, because as I mentioned back in April, we’ll likely have two versions of this instrument. The complete version will be expensive. There is soooo much more content in this than in Sunset Strings, I need to price this much higher, which will probably be out of reach for a lot of people, so we’ll also want a Lite version. But how light is Lite? Do we leave out a lot and sell it at $99? Or do we make the Lite version almost everything except the Legatos and sell it at $299 ($199 on sale)? That will take some time to figure out pricing for both.

No worries, though. With a little luck, we’ll have it all sorted by NAMM 2025! ;)
 
Mike Greene said:
OUR STORY CONTINUES!

Back when I wrote my April update), it was embarrassing enough to concede I wouldn’t be finished by NAMM, but I at least figured we’d be done before summer, along with a triumphant new entry in this thread. Well, that didn’t happen, so let’s make that end of summer. Oh, did I say end of summer? Let’s make that before Black Friday. For sure!

Ah, those days when I was young and believed silly things. Now that I’m older and wiser, I’ve set a more realistic goal of having this ready by NAMM. No way I’ll miss that deadline! Wait a minute, NAMM is just a couple weeks from now? ~sigh~

SO WHAT’S BEEN THE HOLDUP?

Before I get to the stuff you actually came to thread for (the “how to build a library” part of our story), “life” has a way of complicating things. Nothing bad, mind you, but after decades of chasing career goals, and prioritizing work over other stuff (although can making music and having fun creating libraries really be considered “work”?), I came to realize I need to lead a more balanced life. Or more accurately, my wife instructed me that I need to lead a more balanced life.

So we’ve been doing a lot of traveling, theater, events … if she’s bookin’, I’m goin’. And she’s right, it’s nice. (Two and a half weeks in Ireland was great!) Just between us, though, can I admit that when we’re listening to musicians in a pub, or at Disney Hall, I sometimes start thinking about what articulations would be needed to recreate what we’re hearing? Yeah, you know what I’m talking about, right?

Another factor is we started a studio renovation this summer. Back in 1993, I scrounged up every nickel I had, then borrowed a whole bunch more nickels, and built this place. (By “built,” I should clarify that this was a 3,000 square foot warehouse and I built out the interior.) It was a big risk, but I was producing records at the time, and realized I could double-bill the record companies by collecting a producer fee plus charge for the studio, too. Greed, thy name is Mike Greene!

Certain choices in decor I made back then haven’t aged well, though. Plus flooring isn’t supposed to last 30 years, so I consented to hiring a decorator and we’ve been re-doing the whole place. New paint, flooring, cabinets, lighting, artwork … might as well do it all. That way when friends come over, I won’t have to keep making excuses for, “Here’s what happens when you put an adolescent straight male in charge of decorating.” (I don’t care what my wife says, I thought it all looked great at the time!)

What a massive undertaking this is. For starters, there was way so much junk here that creeps up on you. Things that need saving for a few years are a little less critical after a couple decades. (I sure hope nobody calls to ask if I still have their old 2” tapes!) I have to stay involved in all that, because otherwise I can only imagine hearing, “Hey Mike, you had too many mics, so we tossed the old ones.”

It’s been a few months, floors and paint are all done, and now we’re on fixtures and furniture, but I had to put it on pause, since it was stressing me out too much. Especially as it drove me insane that I was spending so little time on the choir project. I do have talented people working with me, but this was a perfect illustration of the negative side of “leading by example.” In our weekly Zoom meeting, I’d too often say “I didn’t get a chance to work on anything this week,” and that became contagious.

So that’s part of the delay. (Although reading it back, it might be a little overstated. Those things took a lot of time, but not all my time.) The biggest part of our delay story is …

WHY, OH WHY DID WE DECIDE TO INCLUDE LEGATO???

We recorded legatos, but there was still a question as to whether to actually include them. Now, I know what you’re thinking: “But Mike, that’s crazy! If you already recorded it, why wouldn’t you include it?”

Two reasons. First is that until we actually test it, we don’t know if it will actually sound good. You don’t want to release a library with weak legato, since many people will judge a library entirely on that. And many of those people will make YouTube videos, or post on forums, tossing around words like “unusable.” (You know what I’m talking about, right? ;) )

So until you know a major feature like legato will actually sound good, it’s best not to promise anything. Too risky. Heck, there’s no legato at all in Sunset Strings and it’s selling just fine.

The second concern is a bigger one. That is that the legato, at least as I’ve designed it for this library, is a massive editing task. I mean massive. Not just in the chopping up of samples, but also in the tone and level matching. I want this legato to be really, really good, so there’s a lot that goes into that.

Rewinding a bit, I’ve done legato before with Realivox. That worked out well, I believe mostly because I spent a lot of time volume balancing the samples. A volume (or tonal) mismatch in a legato transition sticks out like a sore thumb, and IMO, most libraries with reputations of weak legato could improve themselves by just going back and doing some level matching. Easier said than done, though, because that means going through each sample, one at a time, listening to how its volume compares to its starting and landing counterpart samples. It’s very time consuming.

It’s also not something just any sample editor can do, so this part of the editing process is delegated to the A-Team. Which is also the more expensive team. Truth be told, I can’t resist doing some of the editing myself, especially in the beginning stages, where I want to hone in on what the best process would be.

continues ....
Click to expand...
I love these updates and especially all the details you include. It really helps me understand the challenges of putting together a sample library.

I’m also very much looking forward to this one!
 
Mike Greene said:
I had hoped to have an “In Progress” version for the people who want to buy it now, similar to what we did with Sunset Strings and Nightfall, where you could buy the mostly finished unencoded version, then you’d get the completed KPlayer version (for free) when we released it. Essentially a pre-sale, with the added bonus of getting an “in progress” version early. That worked really well.
Click to expand...
Glad that went well! Shame it won't work here, but if it's ever a consideration for future products I'd even be happy buying a pre-release version and still needing to pay a minor download fee for the final release (or other mid-development updates). Not sure what ecomm solution you are on, but some make it easy.

Some people call it paying to be a beta tester. I call it paying for early access.

EDIT: Being clear, I'm good paying more for a pre-release version than getting a discount.
 
Last edited:
Mike Greene said:
SO … WHAT EXACTLY IS THIS PROCESS?

I’ve been debating with myself whether to share this part, since there’s a bit of secret sauce to it. But then I started thinking the real “secret sauce” isn’t so much the “how,” but rather it’s the “time,” and the fact that I’m willing to actually do it. (Or rather that I’m willing to pay people to do it. At A-Team rates.)

In fact, there have been times I’ve shown other developers how I put together Realivox Blue and they’ve said, “I’d never have the patience for that.” That’s not a knock on them that they choose other ways to spend their time, mind you. This is the sample library business, not the sample library lets-hope-our-ASCAP-checks-are-enough-to-cover-our-losses-because-we-spent-so-damn-much-time-on-this business. Honestly, from a pure dollars and cents business perspective, I’m not so sure this legato thing was such a good idea. We’ll see. (Spoiler alert, it’s sounding great, so in answer to my earlier concern, we will indeed include it. But will it result in selling more copies than what the legato cost? I don’t know.)

Anyway, with Realivox (both Blue and the Ladies), the recorded legato intervals are just a 2nd up, a 4th up, a 2nd down, and a 4th down. (Surprise!) All the other intervals are covered with just those samples. (And some fancy scripting work. Sometimes I impress even myself!) And it worked! Or at least it's “good enough,” since I have yet to hear anyone complain that the legato samples don’t sound genuine across all the chromatic intervals.

Two things, though. Realivox is solo voice. And it’s dry solo voice. You can’t use these scripting tricks with an ensemble (for starters, they don’t all slide at exactly the same time), and especially not an ensemble recorded in a large space, since telltale room ambience from the initial note bleeds into the transition.

So for Sunset Choir, we had to go chromatic, an octave’s worth of samples in each direction. That’s a LOT of samples! Here’s the math:

1. 12 up intervals + 12 down intervals = 24 legato samples for each note.
2. For the sustain samples, most choir libraries are recorded in whole steps, so we decided to do the same. (Realivox Blue/Ladies/Men are chromatic, but if the big boys are doing whole steps, and if recording time is costing me thousands per hour, then by golly, whole steps it is!) So that’s 15 “landing notes” to fill the range. So now we’re at 24 x 15 = 360.
3. Men and Women recorded separately. (360 x 2 = 720)
4. Three speeds: regular legato, slow bend legato, and slower bend legato. (720 x 3 = 2160)
5. Two dynamics (mp and mf) for the regular legatos: (2160 + 720 = 2,880)
6. Three vowels for all of them: Oo, Ah, and Mm. (2,880 x 3 = 8,640)

I don’t know about you, but where I come from, 8,640 is a lot of samples. All of which need to be edited separately, sample by sample. And when I say “edited,” the chopping part is just step 1. Here’s a rough explanation of the full process:

1. We recorded our legatos in a similar method as what I think Jasper does (although I’m not precisely sure about his process), where I had them sing these various melodies I wrote. Four of these “melodies,” which, when you deconstruct them, include all the intervals we need. (It’s like that typing exercise: The quick brown fox jumped over the lazy dog.)

2. Editor #1 chops those performances into short snippets of all the legato intervals, then sorts them into folders for each landing/destination note. (The “landing note” is the note a legato interval lands on.) Inside that folder, the snippets are sorted lowest to highest.

3. Each vowel/sex/dynamic/speed combination gets its own Pro Tools session. There are 24 tracks. (72 tracks in reality, since there are three mic positions, but let’s call it 24 so it’s easier to understand.) The first track is “Up 12” (12 half steps = full octave) the second track is “Up 11” (11 half steps = Major 7th) and so on, with the last track being “Down 12.” (Actually, there are more than that, since some intervals got duplicated across those four melodies I wrote. So for example, we have two “Down 12” tracks to choose from. In reality, we have 29 tracks.)

The landing pitches for all the snippets are laid out with lowest note at 0:00 on the Pro Tools timeline, next note at 0:20, then 0:40, then going up the scale to the highest note as we keep moving up the session timeline. So if you think of this PT session as a grid, the landing note pitches go left to right, and the intervals go bottom to top.

Now that we have a Pro Tools session filled with all these legato snippets, we then add the sustain samples, so the legato snippets can be stitched (crossfaded) into the sustain samples. Editor 1 then raises or lowers the volume level of each legato snippet so it matches the sustain sample it will be blended into.

4. Editor #2 (this is higher level work) will crossfade the legato snippets into the sustain samples. First he adjusts the volume of the legato snippet so that the landing section matches the volume of the sustain sample. (Editor 1 got it close, which saves preliminary time for Editor 2. We want the volume-matching end result to be as close as possible.)

4b. He may also need to adjust tone/timbre of the legato, if it's not a perfect match to the sustain. The legatos were obviously recorded separately from the sustains, and I tried to get them to match timbres from one recording to the next, but with vocals, “Sing it exactly like you did yesterday” is never going to be perfect. Especially when they start singing these legato melodies, where with those brilliant masterpieces I wrote, how could they not get carried away with emotion!

4c. I should mention that earlier in the process (between steps 1 and 2), we tuned all the landing notes in those melodies. Otherwise there will be a audible pitch jump when it crossfades into the sustain. (You don’t want to merge a sample that’s 20 cents flat into a sample that’s at 0 cents.) That tuning process is ultra-tricky. (I delayed sharing that step until now since it would have been harder to understand earlier.)

5. Next, he’ll adjust the transition points and find whatever crossfade sounds smoothest. This is by ear, and it’s not a fast process.

6. We use “Consolidate Region” to create new samples which combine the legato snippet with the sustain sample. With some libraries, the legato samples are separate from the sustain samples they go into, but I wanted to eliminate all potential for mismatches (volume jumps or whatever), so I do the combining here, so the legato samples includes the sustains. Note that the sustain volumes with these legato samples are identical to the sustain volumes in the regular sustain groups, so everything in the final instrument should play smoothly.

7. Map the samples into Kontakt and marvel at what a great job we did.

That’s a lot of steps, right? But … uh-oh, we ain’t done yet.

8. We discover a few surprises when everything is mapped into Kontakt. The main one being … well, this one is hard to explain, so consider this example:
* We have a women F4 sustain.
* We have a women F3 sustain.
* We want to play legato from F4 to F3, so we hear the F4 sustain sample, then we hear a legato sample that goes down from F4 to F3.
*The F3 sustain part of the legato sample perfectly matches what a standalone F3 sustain sample would sound like. (Which it should, since they’re the same recording. Remember, we stitched the F4-F3 legato onto the F3 sustain in Pro Tools.)
* But the F4 part of the legato sample is louder than the sustain F4 sample. Uh oh! That means if you play an F4 into an F3, the legato has a big volume bump at the start of the transition. Yikes!

Truth be told, we kinda expected this. A dirty little secret of choir libraries is that there’s usually some volume balancing going on across the scale, since, especially with women, they sing their high notes much louder than the low notes. (Good luck getting sopranos to sing a high A at mp!) So for the sustains, we lower the volume of the high note samples, since it makes for a much more playable instrument. Purists may be offended, but unlike a live choir, which has ways to compensate for imbalances like this, we’re dealing with a sample library, which is less flexible. So the bottom line is do you want “pure” or do you want “sounds good”? (Rhetorical question. You’ll get what we give you, dammit!)

So we need to tweak (volume curve) those particular legato samples so that the transitions sound smooth. It’s an annoying extra step, but fortunately this is only an issue for intervals close to an octave. (Intervals up to a fifth or so are fine.) So that takes us to our final step:

9. Load those tweaked legato samples into Kontakt and now marvel at what geniuses we are!

Great! You've completed one set! Now repeat that process for all the other vowels, dynamics, men/women and speed combinations. Don't you wish you could be an editor and have all that fun?

... continues ...
Click to expand...
Just skimming this made my wrist ache (in a good way though)
 
Mike Greene said:
I’ll need more than Sunset Choir to make 2024 a big year, though. As fate would have it, when it started looking impossible for Sunset Choir to be released before Black Friday, I decided to dive back into the Realivox Men, thinking it might be doable to release the Men before Black Friday. (You really, really, really want to have released something new in the year before Black Friday.) Turns out that wasn’t doable, either, but I did make major progress on it before throwing in the towel when I realized the Men couldn’t make the deadline, either.
Click to expand...
So happy to hear that the Men are likely to make an appearance this year as well!
Mike Greene said:
I had hoped to have an “In Progress” version for the people who want to buy it now, similar to what we did with Sunset Strings and Nightfall, where you could buy the mostly finished unencoded version, then you’d get the completed KPlayer version (for free) when we released it. Essentially a pre-sale, with the added bonus of getting an “in progress” version early. That worked really well.
Click to expand...
I really hope you change your mind on this—and decide to do it for both libraries. At least for the full expensive version. I’m most interested in getting hold of them sooner rather than later.
 
Mike Greene said:
I had hoped to have an “In Progress” version at the show for the people who want to buy it now, similar to what we did with Sunset Strings and Nightfall, where you could buy the mostly finished unencoded version, then you’d get the completed KPlayer version (for free) when we released it a few months later. Essentially a pre-sale, with the added bonus of getting an “in progress” version early. That worked really well.

I can’t do that here, though, because ...
Click to expand...
Evans said:
Glad that went well! Shame it won't work here, but if it's ever a consideration for future products I'd even be happy buying a pre-release version and still needing to pay a minor download fee for the final release (or other mid-development updates). Not sure what ecomm solution you are on, but some make it easy.
Click to expand...
jbuhler said:
I really hope you change your mind on this—and decide to do it for both libraries. At least for the full expensive version. I’m most interested in getting hold of them sooner rather than later.
Click to expand...
I should have said we can't do that in time for NAMM. Mostly because I won't know pricing by then. Plus I won't know by then what will be included in the full version and what goes into the Lite version. With that latter issue, there's the potential for a lot of angry customers if they thought they were going to get X, Y and Z in the Lite version, but we later decide they only get X and Y. And will the Lite version be $99? Or is it $199, in which case it should be substantially less "lite" in terms of content.

There are ways we could make all that work before NAMM, but it's more mental effort than I can spend right now. Strange as this may sound, I just don't know enough yet about what we actually have. In fact, the only playable version of this instrument (other than legato or other one-off tests) is the version I had at NAMM in April. Which ... might be the same version I show this year, since there's only so much you can do in a 5-10 minute demo anyway.

With all that said, we probably will do a pre-release version for the choir, like we did with Sunset and Nightfall. Hopefully soon after NAMM. Having been a customer myself, I can certainly understand wanting an early version, even if it's incomplete, so I'll definitely prioritize making that work. (And I suppose getting money sooner doesn't hurt. :grin:) The nice thing about doing it is that most people already know the drill - "Some stuff will be missing or isn't going to work right, but it should be good in a couple months."
 
Markrs said:
hey why not go for a lite (less features and no legato), regular (all features and no legato) and pro (everything) and then you get to mop up all the potential customers 😂
Click to expand...
I think this was discussed pre-legato announcement (if we're calling Mike's recent news an "announcement" lol) but I genuinely think there's some validity to what you're saying.
I don't know if 2 versions or 3 versions is better, but I do agree that when struggling to price a library like this, tiers are an effective way to "mop up all potential customers"!
Though that's probably paying for NI encoding twice...
 
Mike Greene said:
SO … WHAT EXACTLY IS THIS PROCESS?

I’ve been debating with myself whether to share this part, since there’s a bit of secret sauce to it. But then I started thinking the real “secret sauce” isn’t so much the “how,” but rather it’s the “time,” and the fact that I’m willing to actually do it. (Or rather that I’m willing to pay people to do it. At A-Team rates.)

In fact, there have been times I’ve shown other developers how I put together Realivox Blue and they’ve said, “I’d never have the patience for that.” That’s not a knock on them that they choose other ways to spend their time, mind you. This is the sample library business, not the sample library lets-hope-our-ASCAP-checks-are-enough-to-cover-our-losses-because-we-spent-so-damn-much-time-on-this business. Honestly, from a pure dollars and cents business perspective, I’m not so sure this legato thing was such a good idea. We’ll see. (Spoiler alert, it’s sounding great, so in answer to my earlier concern, we will indeed include it. But will it result in selling more copies than what the legato cost? I don’t know.)

Anyway, with Realivox (both Blue and the Ladies), the recorded legato intervals are just a 2nd up, a 4th up, a 2nd down, and a 4th down. (Surprise!) All the other intervals are covered with just those samples. (And some fancy scripting work. Sometimes I impress even myself!) And it worked! Or at least it's “good enough,” since I have yet to hear anyone complain that the legato samples don’t sound genuine across all the chromatic intervals.

Two things, though. Realivox is solo voice. And it’s dry solo voice. You can’t use these scripting tricks with an ensemble (for starters, they don’t all slide at exactly the same time), and especially not an ensemble recorded in a large space, since telltale room ambience from the initial note bleeds into the transition.

So for Sunset Choir, we had to go chromatic, an octave’s worth of samples in each direction. That’s a LOT of samples! Here’s the math:

1. 12 up intervals + 12 down intervals = 24 legato samples for each note.
2. For the sustain samples, most choir libraries are recorded in whole steps, so we decided to do the same. (Realivox Blue/Ladies/Men are chromatic, but if the big boys are doing whole steps, and if recording time is costing me thousands per hour, then by golly, whole steps it is!) So that’s 15 “landing notes” to fill the range. So now we’re at 24 x 15 = 360.
3. Men and Women recorded separately. (360 x 2 = 720)
4. Three speeds: regular legato, slow bend legato, and slower bend legato. (720 x 3 = 2160)
5. Two dynamics (mp and mf) for the regular legatos: (2160 + 720 = 2,880)
6. Three vowels for all of them: Oo, Ah, and Mm. (2,880 x 3 = 8,640)

I don’t know about you, but where I come from, 8,640 is a lot of samples. All of which need to be edited separately, sample by sample. And when I say “edited,” the chopping part is just step 1. Here’s a rough explanation of the full process:

1. We recorded our legatos in a similar method as what I think Jasper does (although I’m not precisely sure about his process), where I had them sing these various melodies I wrote. Four of these “melodies,” which, when you deconstruct them, include all the intervals we need. (It’s like that typing exercise: The quick brown fox jumped over the lazy dog.)

2. Editor #1 chops those performances into short snippets of all the legato intervals, then sorts them into folders for each landing/destination note. (The “landing note” is the note a legato interval lands on.) Inside that folder, the snippets are sorted lowest to highest.

3. Each vowel/sex/dynamic/speed combination gets its own Pro Tools session. There are 24 tracks. (72 tracks in reality, since there are three mic positions, but let’s call it 24 so it’s easier to understand.) The first track is “Up 12” (12 half steps = full octave) the second track is “Up 11” (11 half steps = Major 7th) and so on, with the last track being “Down 12.” (Actually, there are more than that, since some intervals got duplicated across those four melodies I wrote. So for example, we have two “Down 12” tracks to choose from. In reality, we have 29 tracks.)

The landing pitches for all the snippets are laid out with lowest note at 0:00 on the Pro Tools timeline, next note at 0:20, then 0:40, then going up the scale to the highest note as we keep moving up the session timeline. So if you think of this PT session as a grid, the landing note pitches go left to right, and the intervals go bottom to top.

Now that we have a Pro Tools session filled with all these legato snippets, we then add the sustain samples, so the legato snippets can be stitched (crossfaded) into the sustain samples. Editor 1 then raises or lowers the volume level of each legato snippet so it matches the sustain sample it will be blended into.

4. Editor #2 (this is higher level work) will crossfade the legato snippets into the sustain samples. First he adjusts the volume of the legato snippet so that the landing section matches the volume of the sustain sample. (Editor 1 got it close, which saves preliminary time for Editor 2. We want the volume-matching end result to be as close as possible.)

4b. He may also need to adjust tone/timbre of the legato, if it's not a perfect match to the sustain. The legatos were obviously recorded separately from the sustains, and I tried to get them to match timbres from one recording to the next, but with vocals, “Sing it exactly like you did yesterday” is never going to be perfect. Especially when they start singing these legato melodies, where with those brilliant masterpieces I wrote, how could they not get carried away with emotion!

4c. I should mention that earlier in the process (between steps 1 and 2), we tuned all the landing notes in those melodies. Otherwise there will be a audible pitch jump when it crossfades into the sustain. (You don’t want to merge a sample that’s 20 cents flat into a sample that’s at 0 cents.) That tuning process is ultra-tricky. (I delayed sharing that step until now since it would have been harder to understand earlier.)

5. Next, he’ll adjust the transition points and find whatever crossfade sounds smoothest. This is by ear, and it’s not a fast process.

6. We use “Consolidate Region” to create new samples which combine the legato snippet with the sustain sample. With some libraries, the legato samples are separate from the sustain samples they go into, but I wanted to eliminate all potential for mismatches (volume jumps or whatever), so I do the combining here, so the legato samples includes the sustains. Note that the sustain volumes with these legato samples are identical to the sustain volumes in the regular sustain groups, so everything in the final instrument should play smoothly.

7. Map the samples into Kontakt and marvel at what a great job we did.

That’s a lot of steps, right? But … uh-oh, we ain’t done yet.

8. We discover a few surprises when everything is mapped into Kontakt. The main one being … well, this one is hard to explain, so consider this example:
* We have a women F4 sustain.
* We have a women F3 sustain.
* We want to play legato from F4 to F3, so we hear the F4 sustain sample, then we hear a legato sample that goes down from F4 to F3.
*The F3 sustain part of the legato sample perfectly matches what a standalone F3 sustain sample would sound like. (Which it should, since they’re the same recording. Remember, we stitched the F4-F3 legato onto the F3 sustain in Pro Tools.)
* But the F4 part of the legato sample is louder than the sustain F4 sample. Uh oh! That means if you play an F4 into an F3, the legato has a big volume bump at the start of the transition. Yikes!

Truth be told, we kinda expected this. A dirty little secret of choir libraries is that there’s usually some volume balancing going on across the scale, since, especially with women, they sing their high notes much louder than the low notes. (Good luck getting sopranos to sing a high A at mp!) So for the sustains, we lower the volume of the high note samples, since it makes for a much more playable instrument. Purists may be offended, but unlike a live choir, which has ways to compensate for imbalances like this, we’re dealing with a sample library, which is less flexible. So the bottom line is do you want “pure” or do you want “sounds good”? (Rhetorical question. You’ll get what we give you, dammit!)

So we need to tweak (volume curve) those particular legato samples so that the transitions sound smooth. It’s an annoying extra step, but fortunately this is only an issue for intervals close to an octave. (Intervals up to a fifth or so are fine.) So that takes us to our final step:

9. Load those tweaked legato samples into Kontakt and now marvel at what geniuses we are!

Great! You've completed one set! Now repeat that process for all the other vowels, dynamics, men/women and speed combinations. Don't you wish you could be an editor and have all that fun?

... continues ...
Click to expand...
Honestly, it seems like some miracle that every string library these days isn't priced like Spitfire's Abbey Road Strings.

After reading this...I never want to buy a sample library on sale ever again.
 
I am getting overwhelmed reading this. I mean, yeah, anyone could handle 6000 samples. Easy! But over 8,000? Yeah, that is a lot.

6000 is too much too. I get overwhelm at 20 tracks of vocals on a song. Sigh.

This may end up being above my spending limit for choir libraries I never use, unfortunately. But I am really looking forward to hearing the end result. (as a note, I rarely use choir libraries - I just like the sound of them) I am looking forward to the Men though.
 
dzilizzi said:
I am getting overwhelmed reading this. I mean, yeah, anyone could handle 6000 samples. Easy! But over 8,000? Yeah, that is a lot.
Click to expand...
I remember having only one sample, 1/10 of a second long, spread over 88 keys, aliasing old school
 

Attachments

  • Screen Shot 2024-01-09 at 9.47.22 AM.png
    Screen Shot 2024-01-09 at 9.47.22 AM.png
    44 KB · Views: 16
You must log in or register to reply here.
Top Bottom