Your mention of the flute reminds me of the question of dynamic range, which is one of the most significant factors in making a sample performance sound natural. In fact, dynamics either horizontally with expression, or vertically with dynamic range, probably do more than anything else to make samples sound good or bad. You can take great sample recordings and make them sound absolutely miserable by ignoring what happens in the orchestra. For example, the way woodwinds simply disappear in a Bruckner symphony during the massive fortissimo climaxes. A sample user might think "I've got to allow them to still be heard at least a little" but actually, the oboes, flutes, clarinets, bassoons - they are simply gone in many of these situations because they physically cannot exist next to the combined brass fortissimo. They might have well have exited the stage (though it would be a bit distracting to the audience). And this same "drawback" of actually losing sounds happens all the time in other pieces to a less extreme degree. I think that when one is doing a sample performance, one never wants to lose any of the precious notes one played into the computer, tweaked lovingly, etc., etc. And so there is an extreme unnaturalness that comes with totally distorting the dynamic range.

Another aspect of this is in timbral variation of dynamics - in a live orchestra, even if there is a general mf, a soloistic line in the flute might be played at a near-forte, rather than a p as marked in the score, simply IN ORDER TO BE HEARD. And yet it is played delicately, to give the feeling of p. This happens constantly, in many variations, and all of it impacts samples tremendously.

Another thing I've often noticed in relation to dynamics - listen to a line in a live orchestra marked ff. You would naturally think, "well, I must use all ff samples." No! Because those samples were recorded with a nearly obsessive focus on the part of the player and conductor to get a "pure" ff. This NEVER happens in a live situation. There are constant little dim, cresc, mf, fff, f, variations. So if you use all those perfect ff samples, it is totally artificial sounding.

I guess this is a little off topic of questions of pure sound and reverb, etc., but I think this dynamic question affects the overall impression of the sound of the orchestra more than people generally imagine. If you are hearing instruments at a uniform level which contradicts their natural imbalance, it is as if you are no longer in a natural envorionment. so the most perfect reverb in the world will not help at all. And you might conclude that incorporating these "flaws" - not ony in the sense of dynamic range and expressive and unwritten variations, but also extreme mis-timing, even out-of-tune playing - are the entire basis of natural sounding use of samples.

@William said:

[..] This NEVER happens in a live situation. There are constant little dim, cresc, mf, fff, f, variations. So if you use all those perfect ff samples, it is totally artificial sounding. [...]

Good point. That's why the best virtual orchestrations are those which make extensive use of the huge amount articulations and performance elements in our library.

@William said:

[..] This NEVER happens in a live situation. There are constant little dim, cresc, mf, fff, f, variations. So if you use all those perfect ff samples, it is totally artificial sounding. [...]

Good point. That's why the best virtual orchestrations are those which make extensive use of the huge amount articulations and performance elements in our library.

This brings a lot of threads full circle-- those which may in turn seem OT--

and that is VI farms- in order to run all the articulations at once needed to control and manipulate the wide range of articulations, subtle dynamic ranges-- and any ambience processing which requires serious enough CPU and RAM that it could otherwise constantly compete with resoruces needed to run the samples. One tweak on the reverb and my whole front end chokes.

This means that running one instrument or a group of instruments on one computer-- (or each computer as resoruces allow)-- out as audio tracks for the sake of saving resources-- is going to make doing a truly proper mix a nightmare. Fine tuning real time expression against real time ambience processing remains quite awkward. Real time mixing for the vast array of available samples is unavoidable for the best results, really. Otherwise, one's entire project stands in danger of one compromise or another. Some things are still quite out of reach.

I knew that a farm network would be necessary, but I am constantly surprised at how soon I hit a brick wall, only to find that I need to invest another $2-3k here and there to enable processes that are necessary to acheive the disired result. The 24-instance stress tests results on one computer were encouraging, but not realistic.

So, even though these issues appear to be quite OT, we're recognizing that they are indeed quite closely connected. In order to pull off the required complexities to incorporate the best the library has to offer, plus to have the sort of virtual ambience required to create convincing requirements requires serious cash. The collection can be run with less gear, but that alone makes Dietz's advice difficult to implement.

For me, this only brings greater gravity to the need to 64-bit addressing a lot sooner than developers appear willing or able to offer at this time. The reasons why faster addressing is not available now are unimportant to me in the context of this discussion-- understanding why we don't have 64-bit addressing doesn't help overcome any of today's current hurdles . Honestly, had I known that running the Cube would require an investment of 2-3 times the cost of the libary itself, I might have hestitated to buy it. I am glad that I have the sounds, even if I can only use them in smaller instrument groups and not as a full-on orchestra. I can *almost* get the strings I need on one computer.

This also resurrects the discussion about having violins and violas on different licenses. The reasons why this is not possible are understood, but they don't help the user find another solution-- no amount of money invested can overcome this hurdle, so making the most of the variety of articulations VSL offers becomes not only improbable but impossible.

So, in order to even address the ambience processing, some of the samples have to be eliminated. Even Jerome stated with his 8 Mac Mini setup that certain types samples had to be omitted from his template. 8 Mac Minis plus a front end tower-- and samples must still be eliminated. There's something wrong with that....

At some point, things go quickly from being merely sophisticated and complex to being impractical or unaffordable or impossible. I'm a long way from working the way I need to work with the speed that goes along with it. Load times are still too slow, Syncrosoft is still a pain, processing has to be used frugally at the expense of musical complexities. These and other hurdles serve as severe distractions to making the most of the Cube...

... but I'm working on it.

One favorite quote from Tolkien that applies here:

"... though I buy it with great pain, it is precious to me..."

How about adapting to what's already there?

@Guy said:

How about adapting to what's already there?

as I've said-- I'm working on it. I can't do any better than that.

Nice to see some action here. I originally wrote a bloody novella, but edited it considerably, out of compassion for humand-kind. [;)]

So, to save a few hundred words of mere description, here's a clip:

http://www.rubato-music.com/Media/mp3/LvB_phantom_reverb_example.mp3

That snippet is totally dry, from beginning to end, yet it really sounds as though certain sections, or even just notes, have a kind of reverb applied. (And please, no "lessons" about short-note samples, and release samples -- I'm aware of what they are and how they work, and am aware of what I'm hearing. I'm actually just relaying the subjective experience of my client after listening to the passage. Also, I'm aware that it really doesn't make sense to listen to these samples totally dry, and told my client as much when they expressed concerns about the spatial "tricks" they observed.) I broke the passage down, note-by-note, and realized that this is *strictly* the result/effect of hearing the natural decay of the instrument, as it interacts with the early reflections of the room. The most obvious passage is the forte perf-legato toward the end, which seems to jump into an entirely different space. Obviously, we can't have the smoothness, and beautiful naturalness of the perf-legato without this side-effect, so it seems to be a matter of somehow adding such a "releaseverb"(!) to the rest of the articulations. Of course, I could also just choose a different articulation, but the perf-legato seems musically most appropriate, and avoids too much exposure to the perf-marcato, which has a *very* distinct "signature" in its attack portion. This is why I was wondering about tricks to give those notes which are lacking pronounced natural decays more of the release-decay effect. More specifically, I was curious whether the MIR will find ways of negotiating such spatial/resonance side-effects. (For example, if I smoke a sufficient quantity of opium, I can imagine a system whereby the MIR, since it is said to use meta-data about which instrument being played, actually *inserts* these release-like resonances into the music -- almost like adding a form of release sample at the processing stage... but... as I say, I only imagine this after loads of opium!)

It's worth mentioning, though, that these acoustic effects are only an issue because of the fact that the VIs have come astonishingly cloes to duplicating "real" performances. The realism of the VIs is so convincing that our ears become fine-tuned to genuinely "natural" hearing, which is to say prediction. With other libraries much of this would not be an issue, as our ears wouldn't actually be "tricked" for so much as a measure, let alone a whole passage. This is also why Dietz's solution of using more articulations, in a more varied way, is a good one, but can also be problematic -- as the realism of the performance approaches parity with the "real thing", our sensitivity to unexpected acoustical variances becomes proportionately more acute. The only thing we can do is to select samples, and make decisions about *adjacent* articulations, very carefully...

JWL. As far as "solving" the problem of number of articulations vs RAM/CPU/HD usage, I suspect it can be done (and Herb is all-too-aware of my position on this subject!), but I honestly don't think it can be done in real-time, as an instrument. I'd say more, but there's not much point. I know what would be better for my workflow. But that's just me. If your workflow is entirely different from mine, you could arguably manage large orchestral arrangements on a single machine, with 2 GB RAM -- the VI design is probably the only piece of software that makes such a feat genuinely feasible. I've been tempted into listening to demos of other libraries online, and rarely get through so much as a single mp3 (the Garritan violin and Synful are the only exceptions) -- the lack of sampled-legato is a *glaring* omission, and strikes all other libraries off the list, in my opinion, and for my purposes.

Undertaking the creation of the Vis has been an outrageously ambitious probject for the VSL. When we "buy in" we vicariously share in that ambition. But the fact that I am, once again, in the position of actually considering dropping another $3k on hardware, is the result of my ambition alone. The *potential* exhibited by the VIs has poured a bucket of fuel on that fire, but it could never be said to have ignited it in the first place. It was burning long before VSL even existed.

Dietz. Any more thoughts on my "phantom reverb" issue, after hearing the example above?

cheers,

J.

Uh - please bear with me ... I'm not in the vicinity of any serious playback system right now, I'll get back to this tomorrow, Monday. [:)]

Have a nice weekend!

sorry... my brain is now twisting about wildly in it's bony little compartment....

The other thing which cannot be discounted in this is the fact of the disparity between the performed dynamic, and its influence on the amount of resonance/decay with the instrumental body, and the perceived loudness of the normalized samples. When we hear, for example, a p perf-legato played next to an f perf-legato, the perceived amplitude may be minimal (or non-existent, in the case of mod-wheel dynamic switching), but the influence of bow pressure and string vibration on the prominance of the legato-transition, or any release sounds, is profound, thus influencing the phantom reverb I refer to earlier. This is also something that MIR could possibly deal with, by simulating a more realstic dynamic gradation from ppp to fff, but not I suspect without some form of filtering to provide smoother spectral variations with changing dynamic strength. Is such dynamic filtering being included in the MIR?

I'm sure these are questions you can't really answer... But the more I think about it the more I realize that the recording philosophy used for the VSL samples must have taken the MIR into consideration from day-one, and that, just as the VI made questionable aspects of the original EXS/Giga libraries come clear, so too will the MIR make these acoustic questions theoretically transparent. I just wish I had some real sense of the rabbit-hole's depth! [;)]

Come on, Dietz! Give us a little more... just a tiny bit!

J.

@jbm said:

JWL. As far as "solving" the problem of number of articulations vs RAM/CPU/HD usage, I suspect it can be done (and Herb is all-too-aware of my position on this subject!), but I honestly don't think it can be done in real-time, as an instrument. I'd say more, but there's not much point. I know what would be better for my workflow. But that's just me. If your workflow is entirely different from mine, you could arguably manage large orchestral arrangements on a single machine, with 2 GB RAM -- the VI design is probably the only piece of software that makes such a feat genuinely feasible. I've been tempted into listening to demos of other libraries online, and rarely get through so much as a single mp3 (the Garritan violin and Synful are the only exceptions) -- the lack of sampled-legato is a *glaring* omission, and strikes all other libraries off the list, in my opinion, and for my purposes.

Undertaking the creation of the Vis has been an outrageously ambitious probject for the VSL. When we "buy in" we vicariously share in that ambition. But the fact that I am, once again, in the position of actually considering dropping another $3k on hardware, is the result of my ambition alone. The *potential* exhibited by the VIs has poured a bucket of fuel on that fire, but it could never be said to have ignited it in the first place. It was burning long before VSL even existed.

J-- I've tried several approaches, but Dietz rightfully suggested to broaden one's usage of articulations. Without "enough" hardware, I only want to point out the caveats with this as they play against my desire to do so.

I just dropped another $2k on my setup and am likely to go further with another $10k over the next few months. I *am* indeed working on it, knowing that-- like you-- I've bought into more than just the software itself. It's really like being a stockholder where the dividends are the musical results you get.

Your question regarding the release ambience was not directed at me, but I just wanted to say that I know what you mean and hear what you mean. I've not yet gone so far as to route the dryer sounds through a "pre-pre-delay" to compensate for the missing sound of the "silent" room, but your mp3 is typical of my own dry mixes. I have found myself changing samples, or on rare occasion, changing the duration of the notes themselves-- or choosing to live with the result as is. It's a bit of a quandary, indeed.

@Guy said:

How about adapting to what's already there?

as I've said-- I'm working on it. I can't do any better than that.

hehe, just so you know I was saying that in a amicable way, it may of come out rude, wasn't intended.... [:)]

Yeah, sorry JWL. I didn't mean to sound like I was disagreeing with the idea of using more articulations. Obviously, that's the way to get the best final product. Agreed. Cheers.

J.

@Guy said:

How about adapting to what's already there?

as I've said-- I'm working on it. I can't do any better than that.

hehe, just so you know I was saying that in a amicable way, it may of come out rude, wasn't intended.... [:)]

Much appreciated, Guy.

I guess I'm just working so hard to keep ahead of technical demands, but I also get a funny feeling that by the time I get everything going with all needed hardware it will all change! Yikes.

...I just discovered Bidule's "Stereo Squeezer", which seems very good. I continue to be pleasantly surprised by this little app. Nice!

J.

I've had the thought of using a sidechain to duck an early reflection IR matching the silent stage, for those patches where the ambience tail has been snipped off. On paper seems a more elegant solution than layering over more ER's indiscriminately to mask the inconsistencies, or trying to suck out the existing ambience by narrowing the stereo field (for sections).

And this same "drawback" of actually losing sounds happens all the time in other pieces to a less extreme degree. I think that when one is doing a sample performance, one never wants to lose any of the precious notes one played into the computer, tweaked lovingly, etc., etc.

So true!