Oh I was a smart-alec in my reply so it's all good.

I guess I need to re-word one point. I definately don't want automatic performance in one way and I do in another. The humanize timing idea I mentioned would certainly be automatic. Although I think it's not absolutely essential, just like the current humanize, but that it certain would help in realism without manually adjusting all the start times of each pitch...

The thing I don't want automatic is fine-tuning, which I think people misunderstood. I want a very specific sound out of "my orchestra" so I fine tune and I will also take as long as it needs. Any artificial 'clues' revealing the use of samples, I try to get rid of as much as I can... All I meant was that DURING the fine tuning process I would like features that will save as much time and work as possible in getting to that fine-tuned 'right sound'. I doubt anyone would disagree with that, I simply didn't word it well. How to save the amount of fine-tuning time it takes... I am not sure. I gave the crossfade idea, but ultimately just for the goal I just mentioned. I really want fine-tuning time savers more than anything but without diminishing the level of control I have over VSL.

-Sean

Well I’m not sure this is relative to VIP or relative to each individual library, and how the individual samples are edited.

I loaded up some Appassionata violins and alternated between different patches all playing the note E4, and noted the levels in db(RMS and Peak in a Logic Multimeter). 1st at Velocity X-fade 20, then at 110. Patches used are

1 Sus Vib Strong

2 Perf Legato 4 velocities

3 Stacc

4 détaché

5 Pizz.

6 Trem

Vel X-fade 20

1 (sus) -22db

2 (leg) -17db

3 (stacc) -17b

4 -19

5 -25

6 -20

Vel X-fade 110

1 -25db

2 -26

3 -15

4 -24

5 -28

At Vel X-fade 110, when I go from 2 short notes (stacc then detache) the second is 9 db louder. I really can’t believe my ears. This really isn’t acceptable (or am I missing something?). That’s why I’m having to insert automation data note by note. And going from Sus to Legato at low level (Vel X-fade 20) there’s a huge jump, while at Vel 
X-fade 110 it was smooth. Why haven’t all patches at corresponding velocity layers been brought to the same levels?

Let’s forget the pizz and Sfz, but shouldn’t sus, legato, stacc, detache, trem, trill, have very close levels, allowing us to just draw in a bit of phrasing with the Exp-VelXfad faders, and the rest just sounding musically correct?

Can someone from VSL please chime in and maybe tell me where I’m going wrong, or maybe what could be done? [B]

I did some tests yesterday with the Fl. 1. In the mixer I was lowering the sus by 3 db. In some circumstances that can help, but didn’t find any consistent settings that would lessen the need to automate. And since each patch contains 2-4 samples (that we have no access to), there’s not much we can do.

I know that there a tens of thousands of samples that would need to be reedited, but there are programs that should be able to automatically do that.

I would say that in the context of an entire mix, some of these jump do get masked, but when exposed, even through a great reverb, the differences stand out.

I’d be interested to hear how others are getting around this?

Well, I’ve taken the time to do some more testing. Perturbed in thinking of all the time past and future lost just correcting levels with automation.

I looked at the Trumpet in C short articulations, and the variances seems normal (+/- 3db) on several notes that I looked at.

But a extremely bad case is in the Ob II. I tested the note E5 at velocity of 32.

Stacc -24db, Port short -18db, Port med -11db, Port long vib -19db ..... so a 13 db difference in levels. So if you use all those in a phrase, you’ll probably be correcting at least half of them. The jump from -24db to -11db is incredible.

And then surprise, still E5 but up to velocity 110, Stacc -15db, Port short -19, Port short -17db, Port med -14..... so the differences between the articulations are almost inverted with here the stacc being the loudest.

It’s clear that there hasn’t been enough editing done on these samples.

I think this measures are good examples how certain instruments work in real world.

In the higher register of an oboe, very short notes like staccato do have a higher volume range, because the tonal developement with staccatos are not as big compared to longer notes performing in pianissimo. Longer notes develop more tone therefor the volume is usually higher. Further performing vibrato needs more tonal stability, therefor medium long note with vibrato will be always much louder than a staccato in pp.

In ff usually a staccato note will be always louder than a sustained, the whole possible impact is produced in short amount of time. Longer notes, if they are not performed marcatissimo develop the maximum within the note. Further the tone quality is more demanding on a long note compared to a staccato, the musician has to take care that the tone "works".

Generally we take a lot of care in editing our samples, of course we can do mistakes - no question about that.

What we have is a very good reference in our dynamic samples and the dynamic repetition patches for all the mapping values.

If you need more control about dynamic ranges I strongly recommend VI PRO. Beside setting the individual volume of a certain patch in your template, you can adjust also the dynamic range of each patch individually. And at least you can setup volumegraphs on your keyboard to adjust volume for dedicated play ranges, even on single notes within a patch.

best

Herb

@herb said:

very short notes like staccato do have a higher volume range

That's something that I completely over looked. Don't get me wrong, I still stand by my suggestion to find ways to lessen the time it takes to fune-tune to the sound you want... but the inconsistancy between volumes makes a lot of sense now.

With that, I would prefer to keep the range at the maximum as I think restricting the range of an instrument would then produce unnatural results. I suppose something like that has to be handled on a note by note basis, unless you want to restric the ranges or volumes in your template... which as I said might produce an instrument capable of less realistic dynamic flexibility.

Herb (if you're still reading) or anyone else... I have an ultimatum-like question. I'd either want to strongly suggest ways of reducing the time it takes to fine-tune cross fading to get that perfect sound... (and I'm curious what attempts or concepts could possibly be implemented, if any, to help this) OR I would like some feedback on how to get those results quickly.

---------

This is the only example I have already online but here goes... (Just keep in mind, I know a lot more about mastering, reverb, and mix since I made this... and it was in a massive hurry for a class I had last semester so there are plenty of problems with it.)

http://soundcloud.com/iscorefilm/graduation-theme

Below I posted a couple examples of the crossfading used. The french horn example includes more variation with the expression tool instead of the crossfade tool, but the point is that nothing sounded right, like what I imagine a horn player doing, until I drew it like this... and I didn't simply draw it then done, it took listening, drawing, listening, adjusting, a few times to get it the way that sounded right. - The Cello example is simply to show the kind of drawing it takes to get some good phrasing.

I'm not saying that we shouldn't have to fine-tune... but I don't think anyone would disagree that with standard notation there isn't nearly as much detail given to the performer. Cubase is a performance tool, unlike sibelius, so I get that we need to fine-tune... but I need an easier way... and I can't use a simple mod wheel for it, I feel like 'by hand' is too sloppy when in most instances to get a sound that even sounds remotely real I have to do much 'finer' work than what doing by hand would produce.

Any advice on how time can be saved here? Or any features possibilities that could lessen the time it takes to fine-tune? This is the whole reason I started the thread.

Thanks,

-Sean

At 00:21 there is a french horn line

http://img853.imageshack.us/img853/3040/crossfade02.jpg

At 01:44 there is a cello line

http://img854.imageshack.us/img854/3848/crossfade01.jpg

Thanks for you answer Herb.

I must beg to differ though. My specs above reflect short notes. In the real world Stacc - Port short - Port med - Port long have a similar attack shape, they are just shorter or longer, with the portato articulations having of course decays that the stacc doesn’t.  These levels vary enoumously (Stacc -24db and Port med -11db). These portato long samples, do "develope" a bit, but I noted the loudest levels.

You say that "medium long notes with vibrato will be always much louder than a stacc in pp" (I’m not sure that I agree). So if I play a med-long vib note, then a stacc (all in PP) it will sound musical without me editing anything? 

Have you purposely edited stacc louder than the sus at FF? I’ll check and see if that is consistently the case.

A performer can play a passage at FF alternating stacc, marcato, sus, legato, pizz, portato, etc, all at a very even level. Or, in another piece he can vary slightly, and in another there will be even more diferences. As you know, every phrase differs, every context differs.

In the hands of a real musician there are literally millions of articulations because he shapes them according to the phrase, the spirit, and so on. But in the sample world, the articulations are always the morbid same old ones!. In order for me to create music with these samples I need vigorously consistent levels, so that I only have to draw in the "musical phrase and performance" with my automation. Then I can use the great humanize features to introduce "inconsistency", should that add to the context. Instead, far too often I have to spend literally hours with automation, just to bring the samples to a "neural level". 

As I said, I found the short notes on Trumpet in C (I tested 4 different notes at both low and high vel) to being very playable and musical. But many of the WW and strings imo are much less playable.

I may not have totally grasped all the capabilities for VIP (that I use) but I don’t see any way to "correct the levels" in any time saving & consistent fashion. And if there were settings to better the performance and these libraries, I would hope that VSL would furnish them to me.

Can you give us your global editing criteria on what relationship the levels of the different patches  have at VSL? Are legatos always louder than sus? I haven’t detected any consistent pattern in the instruments that I’ve examined closely the last couple days. Do problems arise from the fact that there are sometimes 2, sometimes 3, sometimes 4 different velocity levels?

To really speak well of these matters, we need lots of audio examples. I will be furnishing some in the coming days, and I invite others to do so as well.

All best !

@Another User said:

far too often I have to spend literally hours with automation, just to bring the samples to a "neural level".

I agree with this 100% - Not all the time, but there are often times that to get the sound that best resembles a real player, things need to be neutral... sometimes it doesn't... But I do spend far too much time doing that, and the phrasing as I mentioned in my other post. I want that flexibility, I just don't want to spend hours per phrase to get that 'real' sound.

I realize my comments might be redundant, I guess I just feel like this point isn't being heard very well and that it's absolutely essential to having a realistic and cost-effective workflow using samples.

-Sean

I would love to get my hands on some settings that would lessen editing time. I’ve tried say lowering a stacc patch in a cell by say 3 db. That may "work" at PP velocity leve with regards to a sus patch, but at a FF level it won’t. By reducing the dynamic range of a patch, what is that going to do for me? I’ll probably need all the possible dynamic extremes.

That’s not really what is at hand here though. We’re talking about basic stuff. At velocity 64, playing sus, stacc, & legato, and getting normal results on all instruments.

If there are really workable settings in VIP, I’d love to have it. But if there were settings to make the basic stuff easier, why haven’t we heard about them?

You guys are thinking too much. I understand the wish to save time, but you've been using VSL libraries successfully for many years (I suspect), with these 'transition' faults, without batting an eye, until it became a forum topic. If it was a real issue it would have been addressed a long time ago by VSL. It's an affectation, not an issue. No one but those who listen on a microscopic level hear this, I'm sure. Who do you want to please? The people who listen, or the people who play? If listeners, it's a non-issue; if players, well, that's your ego talking. If yourself, learn to dumb it down like has always been the case.

I am speaking grossly of course, maybe some patches need reworking, but in general, no one hears these slight differences, except the musician who wrote the piece; so move on and think in 8 bit for your listeners.

Shawn

@jammusique said:

I’ll probably need all the possible dynamic extremes

That's just the point though... if an oboist can play a staccato note louder than a sustain... then having a full dynamic range would not allow for a consistant volume level between patches. - One thing I haven't played with is the x-fade on/off option. I would certainly be easier than crossfades in certain instances... but there are still plenty of times where I need cross fading on and I still run into the problem... so it would still be nice to have it addressed.

To VSL (or anyone):

This thread has consisted of two issues, 1) the volume consistancy bit, which has been addressed and 2) the crossfade bit about limiting the amount of work it takes to fine-tune to get a good result... which hasn't properly been addressed. I'm not trying to diminish the first point, but I created the thread entirely for the second point. Please, any answers, feature posibilities, suggestions for users?

Thanks,

-Sean

@Hicks said:

my main concern is the phase on velocity crossfading!

I'm glad you mentioned this. I've been writing and noticing it, then thinking to post that on this thread... but I kept forgetting. I agree completely! Phasing is definitely a concern I have and the solo violin was the exact example I was using a few hours ago.

Could a 'crossfade humanize' feature be possible? Something to automate the imperfections that a real player has... the reason I mention this is because VSL's samples are recorded as plain and even as possible. I certainly get why, there is a need for it in many ways... but it's also in every way inhuman. I still want the ability to fine tune, but the current humanize feature for pitch imperfections allows for both fine tuning and a degree of automatic 'humanization'. Having a similar feature for crossfading, making things more human rather than very flat dynamically in my opinion would be an essential feature.

In the meantime, I agree... VSL still has the best samples available...

-Sean

@Hicks said:

But my main concern is the phase on velocity crossfading! It has been created while ago and since it seems that time haas changed and a lot of improvement have been found. Would it be possible to change the velocity crossfading to try to avoid the most the phase and the "multi" instrument effect (ie when using solo violin and riding the xfading and hearing two violins because samples are layering). Regards

This is a very real concern and I would hope that VSL is actively working on solving it. Phase aligning such a huge sample set is a really daunting task, so i would settle for just the legato and sustain samples. However, this would not suit the people who like to do all dynamic control with a continuous controller. This has always seemed counter-intuitive to me, as "one-shot" samples are perfectly suited to velocity control, but would those people be prepared to accept a compromise?

AFAIK the only sample library developer that allows you to cross-fade on solo instruments without phasing is Samplemodeling, and that is not really a sample library. I would imagine that phase aligning a few MB of samples takes far less time than a few hundred GB. [;)]

DG

@iscorefilm said:

 

Could a 'crossfade humanize' feature be possible? Something to automate the imperfections that a real player has... the reason I mention this is because VSL's samples are recorded as plain and even as possible. I certainly get why, there is a need for it in many ways... but it's also in every way inhuman. I still want the ability to fine tune, but the current humanize feature for pitch imperfections allows for both fine tuning and a degree of automatic 'humanization'. Having a similar feature for crossfading, making things more human rather than very flat dynamically in my opinion would be an essential feature.

I think that for large ensemble patches this is not really necessary, as those imperfections are ironed out when you have many players playing together in the same patch. However, it is useful to have some human element for solo instruments, and I would recommend getting a breath controller so that you can play the expression live. I know some people use a control surface, or even an expression pedal, but particularly for Woodwind and Brass instruments, I wouldn't be without my BC.

DG

I definitely agree with the request for matrixes where the levels of the various articulations are matched.  I see the reasoning behind doing it the other way but that's much less useful for me - if I'm playing a phrase that mixes different lengths of notes, I want to be able to just play the phrase in and have the dynamics reflect how I'm playing it in, not jump all over the place every time I switch articulations.  Seems like the best solution would be to improve the patches as opposed to adding more interface functionality for each user to try and fix it themself.

If there are users who like it the other way, then what about having some sort of switch in the interface?  Or providing two sets of matrixes?

The inconsistent transitions between dynamic layers are an important issue as well, thankfully I have only noticed problems a few times with the VSL libraries.

@DG said:

I would recommend getting a breath controller so that you can play the expression live.

I have one last question for you then. My sole goal is to save time... I often get distracted from writing, specifically 'moving forward' with a piece cause I spend so long working on the getting the beginning sounding right. I'm OCD and if it doesn't sound right I just don't want to move on.

would using a breath controller be faster (more accurate) than using a fader or mod wheel? Because I'm used to drawing in this expression, I do very fine-tuned 'crossfade drawings', so to speak, so I want to know basically if the BC will be better at that than faders, knobs, etc. While those seem faster, they don't allow me to fine-tune quite as accurately.. thus my question.

Thanks,

-Sean