@DG said:

The one big fault in VSL VI is the phasing issue on Velocity xFade. This is the area where I waste most time whilst programming, so if that could be fixed (Herb says not until faster PCs exist) then the next real technological leap has to be much further than anyone has so far suggested.

Just curious, why do we need faster computers to do this? Can they not just "pre-process" the samples? Not every single time, but as a one-off.

Apparently this is what Garritan has done with the violin and cello, so there is no extra strain on the processor trying to perform real-time computations.

Regards,

Ben H
I wouldn't pretend to know what the technical difficulties of this "morphing" are. However, if it sounded like the Garritan instruments, I wouldn't buy it.

DG

@DG said:

I wouldn't pretend to know what the technical difficulties of this "morphing" are. However, if it sounded like the Garritan instruments, I wouldn't buy it.

DG

Hey DG,

I wasn't commenting on the sound of the Garritan instruments so much,

I was more questioning whether some offline, pre-processing of the samples would mean that the heavy computations of "morphing" would not have to be done by the user's processor at all.

Ben H

correct me if i'm wrong, but velocity x-fade would happen in a plane (vel-axis, time-axis) so there would be (almost) endless combinations needed to be preprocessed. and *someone* would need to know where these samples are located and when to use which.
christian

Hey Christian,

I don't pretend to know all the intricacies involved in the process, I was more relaying a response from Gary Garritan when he said that the samples in his violin and cello were "pre-aligned," so that there was no extra CPU hit from calculations for the user.

Regards,

Ben H

I share Ben's puzzlement. The average modwheel crossfade covers four samples per key. Two is not uncommon. Is pre-alignment of four samples (per key) that difficult? It might be. I also don't know anything of the technicalities. To my non-technical mind, it sounds like a tedious and lengthy processing task, but not prohibitive.

I think it is the time element involved (unless I am mistaken). VSL solo violin uses completely real recordings of notes. Garritan uses artificially processed recordings. So you can align those artificially tweaked notes, but you cannot align real, living, human notes quite as well.

This is a bit mysterious. I suspect he's using some variant of additive resynthesis by inverse fft. This method does allow control over phase, if I'm not mistaken. And when you know the precise pitch of each note your resynthesizing, and there's not too much noise content, ifft can sound very good. What I don't get, though, is how this would handle differences/fluctuations in natural vibrato rates between different samples. But maybe it doesn't need to deal with those.... Mind you, I haven't really heard much of the instrument, so I'm not sure how seamless it really is.

J.

@William said:

Garritan uses artificially processed recordings. So you can align those artificially tweaked notes, but you cannot align real, living, human notes quite as well.

Okay thanks.

I did not mean to be critical of VSL, and I am sure that you are doing everything possible to make great, playable, and innovative instruments.

Kind regards,

Ben H

Ben, never be worried about being critical of a product. You are the customer and are paying for the product, so you have a right to expect it to be as good as possible.

Regarding the Garritan instrument, whilst I find the technology very exciting and would like to be able to use it with my VSL sounds, the huge let down for me is the sound. As a violinist I will obviously be more critical of violin samples than I would of others, but I think that it sounds absolutely horrible. I've not heard the raw samples, so I don't know about their quality, but I suspect that some of what sounds so dreadful to me is caused by all the processing that is needed.

The fact that Herb has already mentioned the processing power needed to do xfade properly says to me that at least VSL is trying to come up with solutions. The power of modern CPUs is increasing at a tremendous rate at the moment, and it is possible that in the near future many of the things that we want will be realised. I have a huge list of things that are not good, don't work properly, need re-editing etc. with Vienna VI, but the only reason that I have such a list is that IMO there are no other products on the market that are of suffient quality to know where to start the list. [:D]

DG

I'd say that just about sums it up DG, as usual.

I agree on the sound of the Garritan violin. It sounds to me something like an old-fashioned electric violin. Though not as good. [6]

Also, JBM don't you think what is being done with the aligned waveforms is similar to that "synful" thing? It had a similarly artificial sound, and used a complicated script to morph waveforms for legato and timbral dynamic changes.

OK, folks, here I come: I quite enjoy working with the Garritan violin because you can really play it with full control of vibrato and such. Already before acquiring the Garritan violin I started changing my mind to favor a more musical solution over the perfect sound and was reprogramming all my sounds. I can listen more relaxed to the mockups I did since then.

But I have to agree that most of the demos sound downright horrible. But one can do much better than that.

@mathis said:

OK, folks, here I come: I quite enjoy working with the Garritan violin because you can really play it with full control of vibrato and such. Already before acquiring the Garritan violin I started changing my mind to favor a more musical solution over the perfect sound and was reprogramming all my sounds. I can listen more relaxed to the mockups I did since then.

But I have to agree that most of the demos sound downright horrible. But one can do much better than that.

Mathis, I don't want to derail this thread too much, but do you think that some of the problems with the Garritan violin are connected with the fact that it does have control over things like vibrato? I wonder whether part of what sounds so bad to me, is that I get the feeling that it sounds like a young kid let loose in a sweetshop. Kind of all effect, and no substance. Any thoughts about this?

DG

I simply agree, that's the main reason why the demos sound so.........
Vibrato is a matter of taste and I like to have control over it. (but Strad's vibrato still sucks, of course, it's just controlled suckness...)

Hiring a violin player is so easy that it's pretty useless to really think about it, actually.

I think all those libraries (Ai, BBB, and VI included...) all share the same problem: as they expand and start to allow access to a huge number of articulations, sounds, effects, techniques, etc, they also become more complex.

The "installation" side (getting various computers together, installing the software and the samples, setting up the sequencer...) is complex but only needs to get done once.

The problem is the actual use of this new power by the composers. More parameters mean more controls, which are usually proprietary and thus need to be learned each time a new product comes along.

I really think the problem lies in the fact that we *already* have a language we all speak and commonly use... we have been using it for years... most of us know it extremely well... music notation! And we find ourselves having to learn various technical ways of getting specific sounds out of our various libraries... Even though there's usually a way to convey it easily on a score!

The world of music notation is ruled by strict conventions... this is obviously not the case in the world of orchestral libraries: all of them use different controllers, different interfaces, have different ways of interpreting volume, expression, modwheel, etc. etc.

My dream is for a sequencer with a *real* notation / score program; like a crossover between Logic/DP and Sibelius. You would not use "tracks" for your accoustic instruments anymore, but regular music staves... But also, a notation program that would cleverely analyze its content: When putting a dot over a note, the virtual instrument would switch to staccato. When putting a slur or inputing "legato" it would switch to a legato patch. Etc.

You would be writing music like it's supposed to be written... on a score! And for all the synth, loops, and effects stuff, you would still have regular tracks available.

Notation programs already do that for some things. For example, if you put a "pizz." text element, the program automatically switches to another patch. That's the concept.

I am convinced that the sample libraries in themselves *have* to be huge, with hundreds of thousands of samples, to convey all the necessary articulations and technique.

But the sequencers / notation programs have to catch up at some point. We need some kind of standard API which would allow virtual instruments developers to allow us to use their amazing libraries easily.

When I hear BBB, I'm amazed at the sound quality - but at the same time, I know it will require a lot of tweaking around.

It's just too bad, and it is and will be the biggest issue with great-sounding virtual instruments libraries until "someone" can find the answer on how to integrate these easily with a real notation program.

I spent 20 years learning to play music, and 10 of them learning to write it.... I want to apply that knowledge to the world of synth orchestra... instead of speding time re-learning new tools every other year...

I'm sure that makes some kind of sense... don't you think?

Anyway... sorry for the ranting [:)]

Jerome