Most of the limitations in this area come not from the libraries but the notation programs themselves, and gigastudio (a terrible piece of software IMO, but because it's patented, no one can make a competing product).

Notaiton programs need to respond musically to every articulation, play style, and dynamic symbol that you can put in a score. Siebelius, Finale, none of them actually respond to crescendos, slurs, trills, tremolos, pizzicatos, etc. You'd think a program like Finale would allow more customization in response to different playstyles, but it doesn't allow any except the most basic. Siebelius is the same way. What we need is notation software with the flexibility of a MIDI sequencer.

@Another User said:

"What we need is notation software with the flexibility of a MIDI sequencer."

Yes, I agree using anything less than an extremely professional notation program with a library such as Vienna would be inconsistent with the type of environment we are ultimately attempting to achieve. However, Sibelius can easily be used in such a capacity right now, where one first inputs the notes and simultaneously inputs the expressive inflections with great ease on the staff, so that one never has to leave the score to drive the articulations in GigaStudio.

There is an article describing this process available at http://home.tallships.ca/island/interface.pdf . Some examples of Sibelius used as such are: http://home.tallships.ca/island/hebridesgigaversion.mp3 and http://home.tallships.ca/island/brahms33.mp3

Cheers, Gungnir

Thanks a lot for that info, I'll have to try it. Maybe I've been under utilizing siebelius. I use Finale mostly and know that this is not possible in Finale.

What I would like to see is a program where you can make any symbol appear on the staff and then have total control (perhaps even by programming a macro) over what happens in response to that symbol.

i think that you should not try to program all articulations within the notation programs, even though sibelius can go quite far with hidden ~xyz parameters; actually, i have already suggested them to offer a version with an integrated sequencer backbone, i.e. at least with a list editor and a piano roll, where you can make additions and amendments, yet knowing exactly which elements or parameters have an impact on the notes shown and which are just for playback.

in other words, what we need at this stage is a straight-forward, but very well sounding "basic" set for notation, and then the full library for use with the sequencer: there is no other way to get a good sounding track than the input "by hand" in a good sequencer.

I still say the ultimate notation software would let you author a macro in response to any symbol that you can place on the staff.

IMHO, to achieve a realistic performance, you really have to PLAY each and every line, and then to tweak some parameters. You cannot expect a notation program to perform the way you would like it to perform. It is not human!

So the basic workflow would be to write the score, and then play each line in a sequencer and after that tweak everything.

@Another User said:

So the basic workflow would be to write the score, and then play each line in a sequencer and after that tweak everything.

Wow Martin, are you comfortable that is a true option? Many wish to produce large scale works with the upcoming new libraries, and in my experience with sequencing Brahms symphonies, etc., this would be essentially impossible, even if I had the talent to play the individual lines. I believe the next big step in virtual orchestration is to finally utilize some rather basic artifical intelligence to anticipate what articulations would normally be used within a phrase, and have a notation program insert them in for you (as explained in the Sib-Giga paper). The main melodic lines for espressivo phrases, etc., could be played in directly, but to play in every line in a large work would be gruesome.

The libraries are coming, now we must design our software to easily use them. When Sibelius addresses some timing issues and some limitations with Manuscript, I believe it will provide the best possible solution.

This topic however, along with the issue of ambience, will become a pivotal one when the Vienna becomes available, because it involves the other important cost component, our time.

And this site is the place to discuss it because most people using it are involved with larger scale symphonic renderings. Hopefully an ongoing dialog on the subject will materialize.

Cheers, Gungnir

I've been thinking quite a bit about options for inputing VERY LARGE scores. I am involved in musical theater, creating music scores for theaters that cannot afford live orchestras. It takes two months to input all of the instruments, tweak them, add in live musicians in a couple of judicious places, etc. I am looking for a faster way! Right now, i play in all of the parts, quantize the rhythm section, go back and tweak dynamics, then adust tempos and finally add live musicians and mix. Any ideas on efficiency improvements or macros? I don't use Siebelius or Finale.

I may be totally off, but it seems to me that different players in a large symphonic orchestra (leave out soloists for now) would play the same score the same way with the same conductor. So the real variables are the score (which the software can provide) and the conductor. So maybe we don't need to micromanage every instrument, because we'll assume there are objective ways an orchestra will play things in response to a conductor and a score, but rather work on the composer/conductor's interface with the notation program?

What I mean is, what if there was a way to have a score program with a sophisticated enough interface that you could give broad directions to each instrument group for different parts of the score and they'd follow it, like you were their conductor?

You could easily use the mouse as a baton, and then save the performance for the future or for recording.

I know it couldn't be perfect but it'd be a step in the right direction, no?

Well, I wish you guys were right, but what you are talking about is impossible. A computer cannot emulate something a human would do in regards to something that is artistic. Therefore, if you want to have a musical rendering of a symphonic work, you have no choice but to play each and every line musically, so that the result is musical. I know, I know, this is much too long... But hey, do we really have the choice?

The only thing that can help us is that you only have to play a line once. Right now, for a single violin line, I have to have at least ten tracks in a sequencer to use all the articulations... This is really painful.

I have been a software developper for years, and one thing I know for sure is that a computer will never be able to have an artistic mind. Artificial intelligence is possible, yes, but this stays in the rational world. Things related to musical performance (artistic behavior) are everything but rational. And the only other way a computer can act is randomly... which is not really art...

I truly believe that even in 100 years from now, no computer will be able to give you something musical with just a score.

Martin

@Another User said:

I truly believe that even in 100 years from now, no computer will be able to give you something musical with just a score.

Martin, I agree with this 100%. The performer (even if it's a keyboardist realizing a score with samples) brings an enormous amount of musicianship to a score. It's really a bit naive to discount this. The score is a guide to the performance, and the details need to be filled in by the performer.

It actually strikes me (someone correct me if I'm wrong) that while VSL will be a big step forward toward realizing more realistic performances, it will actually be stepping farther away from the idea of having samples realize a score from a notation program. It seems that the VSL folks have determined (rightly) that many performance gestures (certain phrases, runs, etc.) cannot be realistically performed by simply triggering individual notes. So, they are giving us a library of larger pieces and gestures that can be assembled together to fashion a performance. I would assume (again, someone correct me if I'm wrong) that this is going to require a fair degree of hands-on management (nothing good comes without a price [;)]).

In other words, I don't imagine that VSL would be able to look at a raw score from Finale or Sibelius and automatically and flawlessly insert larger or more specialized performance gestures (e.g. substitute a sampled run for a specific string of notes) when it sees such things in a score. Or will it? My experience with other libraries that have some larger chunks is that significant selection, editing, selection and tweaking needs to be done (with a lot of listening and evaluation from the person wrangling the samples), and the resulting MIDI info usually would have little relationship to information that would come from a score (i.e. the score information, and the performance information needed to realistically realize a score are two separate, distinct entities). Indeed, with some libraries, you first consider the raw elements you have before you decide what *can* be made with them.

Lee Blaske

Peter wrote:

@Another User said:

I may be totally off, but it seems to me that different players in a large symphonic orchestra (leave out soloists for now) would play the same score the same way with the same conductor.

Peter, you must not have ever played in an orchestra. [;)]

It's been my impression that in most orchestras, any given individual section of instruments is convinced that they are playing things 100% correctly, and that every other section is either ahead or behind. OTOH, these disagreements (hopefully minor) often contribute to why a piece realized by a large group of players sounds big.

Lee Blaske

No, never played in an orchestra [:)] So I may be wrong, but I still think that with conductor-like direction from a human, you should be able to get a computer to sound enough like a real orchestra to fool all but experienced orchestral musicians. I don't mean to suggest that it'll ever be a replacement for the real thing (nor would I ever WANT the real thing to be replaced!!) but I think we can get it acceptable enough for most movie and video game scores, as well as off-Broadway theater. Not to mention the wonderful tool these libraries are for the aspiring orchestral composer who no longer needs to have sessions with the orchestra to hear his experimental pieces played.

Plus, of course we'll never capture the differences between the London Symphony Orchestra and the New York Philharmonic, but maybe we could at least capture the Vienna Orchestra and always have our compositions sound like they were played _by_ the Vienna Orchestra?

Will we see a solo viola? How about a second violin?

As a classical trained composer I would really love to use Sibelius for producing my pieces, but I do think that, at least for now, I will say with Logic in order to achieve less mechanical sounding results.

I did visit the sites with the Sibelius-GigaStudio demos of hebrides overture, and I must say that this is not convincing at all to me. I may need to mention that some years ago I was a performer with the Vienna and Berlin Philharmonic orchestra, just to say that I do think having an idea how it sounds in the real world....

By the way I write my music with Sibelius [;)] ....But producing is made with Logic Audio a Mac and 3 Gigastudio PC's.

I am looking forward to the Vienna Librarie

It's probably best to hold off detailed discussion on the topic of notation programs and sequencers until Vienna releases their approach to programming, but I feel inclined to make a few further points on this important topic.

If the only way we will be able to generate large orchestral works using sample libraries is by playing in each line for each instrument, one is almost tempted to forget about the Vienna Symphonic Library, go back to manuscript paper, and hire a real orchestra. The cost of the samples, software and hardware, added to the enormous cost of your time, coupled with the expressively compromised end-result, effectively dilutes the advantages of using virtual orchestration for major works to a critical extent.

Software does what we design to it to do. Nobody will argue that sequencers or notation programs will ever generate the realism in phrasing that humans provide, but the question obviously becomes a matter of degree; how much realism provided at what cost in tools and in time. For example, the sixty bars of Brahms Third Symphony, third movement (mp3 referred to above) takes about three hours to sequence in Sibelius, if one is familiar with their library articulations and if the notes are already loaded. So, with a piece generated by using step input in a notation program rather than by playing the notes directly into a sequencer, you get a more mechanical result but a much faster working period.

And significantly for some, you never have to leave the professional working environment that Sibelius provides. If Sibelius fine-tunes their plugin language, the above sequence would take less than a half hour. Phrases can easily be incorporated into Sibelius without compromising the visual integrity of the score right now, and when virtual orchestration becomes more widely used, some basic plugins could probably do most of this automatically.

It all depends on the style of music you are writing, your realtime playing skills, the length of the work itself and the amount of time you have. But I find it ironic that on one hand technology is about to provide us with the largest sample library ever conceived, and yet some people are stating that the same technology will never be able to provide the means to use it realistically. Very little research and effort has been put into this particular area to date, but I do not agree that computers will not allow extremely convincing interpretations once it is. I believe that Sibelius, due to its intuitive, almost tactile working environment, represents the best hope for the eventual assimilation of a truly professional notation program and playback device.

The Hebrides mp3 was intended to demonstrate the potential of using Sibelius as a sequencer on an "as is" basis, with direct reference to its future development. Nothing was played directly in using Flexitime. It may be interesting for someone to do a version of this piece using the direct input approach, so that an assessment of the relative differences in expressive results, as well as the time required to produce them, could be made. [[;)]]

All the best, Gungnir

[8-)]
Gungnir,

Sorry to not agree with you! It has not to be one way or the other. The Hebrides score is a demo included with the Sibelius software [[;)]] ...and is written in 100%, which garanties a 100% mechanical result. Even with some editing and controlers the rytmic feel will stay like a machine. But it is also possible to mix playing and writing. I even do more mixing because I do play the percussion parts with a keyboard and the string / wind parts with a synthophone-wind controler. Of course I also do write parts in, in case where I do think it will make no difference, sometime even be better or if is to time consuming to play the parts.

Even when the results are not as the real thing, at least for now, there are still lots of reasons to continue using this sample systems. I mean, what nice toys for a musician [:)]

@Another User said:

It all depends on the style of music you are writing, your realtime playing skills, the length of the work itself and the amount of time you have. But I find it ironic that on one hand technology is about to provide us with the largest sample library ever conceived, and yet some people are stating that the same technology will never be able to provide the means to use it realistically. Very little research and effort has been put into this particular area to date, but I do not agree that computers will not allow extremely convincing interpretations once it is. I believe that Sibelius, due to its intuitive, almost tactile working environment, represents the best hope for the eventual assimilation of a truly professional notation program and playback device.

It's not as gloomy as you make it seem, Gungnir. The problem is that an application designed primarily to be a scoring application is not going to be the best hope for musical interpretation because scores, due to their very nature, do not contain an ENORMOUS amount of performance information that we assume performers have. It would be like a word processor with a text to speech function reading poetry.

Now you could conceivably build many hidden layers on top of the notation level of something like Sibelius that could contain all the data needed massage what's down in black and white to actually realize a piece of music. I have a feeling, however, that if Sibelius is ever bloated up to actually do this, it is going to be a very cumbersome, unintuitive piece of software. Regardless, non-realtime tinkering with these aspects is always going to be more painful than what can be accomplished in real time by a performer (even if every line of a large orchestration needs to be played individually). A musical synergy is built up in a real time performance process that is going to be extremely difficult to accomplish in a non-real time way. Personally, I start out with a reduced guide track on a default instrument to establish tempo and basic feel. Then I use that guide in a fixed head graphic editing display window to visually "conduct" the performance (along with the audio). It's amazing how well and fast this works. At some point, I'll lose the guide audio, and I might go back and replace some of the early tracks once more of the orchestration is built up. In this sense, I'm rhythmically and dynamically playing with the orchestra, reacting and relating to all the other "players," hopefully playing in a stylistically correct manner. An enormous amount of intuition and on-the-fly editing comes into play. I can't begin to imagine how this could be accomplished in a non-real time environment with even a great sample library. To begin with, how would you accurately calibrate the dynamics of each solo and section instrument? This is something that's much better done by ear.

Maybe someday we'll have an application that can both score and musically realize the score. Personally, I feel Logic is much closer to that ideal right now. At this point in time, however, achieving *excellence* at either task requires the use of specialized tools. You might as well ask why humans, after thousands of years of civilization and tool making still don't have one single eating implement that simultaneously fulfills the function of knife, fork and spoon. I am open to someone or some company coming up with a solution that elegantly integrates scoring and performance, but it hasn't happened yet. Personally, at its present level of development, I would no more want to use Sibelius or Finale to perform music than I'd want to eat soup with a fork. [;)]

Lee Blaske

Lee,

Great post! This is really exactly what I also do think explained in a very
comprehensive way!

[code:1:3985285a10][i]>Personally, I start out with a reduced guide track on a default instrument >to establish tempo and basic feel. Then I use that guide in a fixed head >graphic editing display window to visually "conduct" the performance >(along with the audio). It's amazing how well and fast this works. At some >point, I'll lose the guide audio, and I might go back and replace some of >the early tracks once more of the orchestration is built up. In this sense, I'm rhythmically and dynamically playing with the orchestra, reacting and >relating to all the other "players," hopefully playing in a stylistically correct >manner. [/i][/code:1:3985285a10]

I do not understand exactly the procedure, but it sounds really interesting. Could you explain once more exactly how you proceed? Sorry my english is not that good!

Thanks

Igor

Great post Lee, I totally agree with you.

I second Igor, and I would also like to understand more the procedure you were talking about. It looks like it is something I have been looking for.

Thanks,
Martin