@gugliel said:

Maybe we can keep this active and get some answers!

One thing more: looking at the wave forms, and listening, it seems that all the keyswitches play the SAME recording -- they just start later and later into the wav file.

So, as James says, maybe the keyswitches step through the recording, a repetition at a time. And since it is all one recording anyway, there is no particular wasted space. Sounds possible! Somewhere in the documentation it was observed that you can use these for FASTER than recorded passages, but that it was inadvisable for SLOWER than recorded passages. This would make sense using them this way.

This may be the case, but from what I saw in the Gigastudio editor, there was actually stored (for each starting note) a full set of 9 sample repetitions all the way down to 1 sample repetitions (obviously the last note in the set). This has confused the heck out of me, because the repetitions seems like they've simply chopped up the 9 repetition set into a full sequence so that you can access the first note:

(the columns are the individually played repetition notes within a single sample, and each row represents a single sample linked to a keyswitch)

1 - 2 - 3 - 4 - 5 - 6 - 7 - 8 - 9
2 - 3 - 4 - 5 - 6 - 7 - 8 - 9
3 - 4 - 5 - 6 - 7 - 8 - 9
4 - 5 - 6 - 7 - 8 - 9
5 - 6 - 7 - 8 - 9
6 - 7 - 8 - 9
7 - 8 - 9
8 - 9
9

However, we still see the full set of nine repetitions, apparently assigned to the nine keyswitches. So, why didn't they chop up and store the sequence like this:

1
2
3
4
5
6
7
8
9

Each note is now assigned to its own keyswitch. Granted, you do lose the ability to play the sequence "naturally". This is the only reason I can think of to store all those sequences (unless I'm mistaken in how the system is operating).

I'm still somewhat confused by this all. Any enlightenment would be appreciated. [:)]

Exactly 25% of both the First Edition and the Pro Edition is dedicated to duplicated samples for the purposes of repetitions. VSL has their reasons, but still, each successive repetition contains exactly the same data as the previous one. So for a 9-rep sample, they have 8! (factorial) redundant samples.

The Pro Edition which wavers around 230GB would only be 170GB if it weren't for the redundant data.

Evan Evans

I'm afraid you are right. Because there is a field called 'sample offset' in gigastudio, I thought the rep instruments would use it to take a different chunk of the same recording for each keyswitch. But the offset appears to have a limit of 2000 samples, which does not take you very far in one of these long recorded repetititions.

The po-rep9_8-130mf instrument in solo strings seems to use two alternating long samples, one on the odd numbers and the other on the even numbers. What a waste of space, both on disk and loaded in memory! I think I'm aiming to do something like the reduced legato project, delete half or 2/3 of the samples and plan to keyswitch in groups of 4 or 5, not 8 or 9!

A simple cut behind each sample part is not the solution.
The next performed note should not be audible, and the fade out has to be as short as possible, to keep the maximum of each note. Maximum fadeout time 5 to 10 MS.
That means, if you perform repetitions slower than the original performance, it would sound extremely artificial, I would say: really bad.
I don't want to imagine the forum topics, if we decided to edit the samples this way.

Keeping the whole sample, makes it more easier: if you are getting doubled attacks in your performance, you know you are performing the repetitions too slow.
There is a clearance, so it will never sound so radical like a hard edited version described above.

Further this solution does not waste any RAM, because streaming only loads small parts of each the samplestart into RAM. The size is much smaller than the datasize of each individual repetition parts.

The number of repetitions is set to 4 for slower repetions (without the starting note), 8 to fast repetitions. We have tested, that less than 8 repetitions does not work for large repetition counts (more than 20 repetions using the chain mode).

Generally each desicion, how the library is designed, is following a concept and are not at random.
Please take the time and read all the manual instructions.

By the way Evan, all this was dicussed several times on this board.
I have no problem to explain these things to new customers, but I hate it to discuss things twice with the same person.
Evan, this is the first of three warnings.

best
Herb

Herb,

Don't warn me buddy. There is no need. This guy and I are desiring other ways to use repetition recordings, other than needing to use 400% (8!/9) of the same data. It's just peaceful discussion.

I fully accept that you have these samples in the set for good reasons. Although I do not understand it. I am sure it is very simple, but no matter how many times you've explained it, I still don't get it. I am a very smart guy ... but for some reason when I place all the concepts, illustrations, and ideas, out in my spatial processing oriented brain, I just can't connect teh dots to see the logical reasoning.

BUT, I still completely understand that there is some reason why we needed the duplicative data in new sample files on the physical hard drive. I just don't know why.

[;)]

Evan Evans

P.S. I understand you don't like explaining things twice, but calm down. You are my equal, a friend, and a fellow person ... NOT Mr. Superior. I cannot be warned by you, nor do I accept your warnings. To do so would be to accept unexcuseable abuse.

Oh, brother, this thread is going downhill fast... Please, everyone. I was seriously just trying to figure out how this stuff all worked.

Herb, I know these decisions aren't made lightly. I'm not accusing you of anything (and I really don't think Evan meant to either). I still don't understand the decision, because it still sounds to me like the same could be achieved by storing the first two samples instead of the entire run. But to be honest, I really don't care all that much. This is obviously something of a hot-button issue between you and Evan, and I seriously don't want to get in the middle of that! [[;)]]

Anyhow, the whole disk space thing was sidetracking my original issue, which was that I kept trying to play repetitions by playing a single note, then adjusting the settings to try to adjust the speed of playback (obviously I didn't have much success). Naturally, it must seem fairly foolish to someone who already understands how it all works.

The other factor that kept me confused was that none of the factory presets for the repetition tool created a convincing repetition effect either (even when played correctly). I'm not sure why all the settings (for First Keyswitch Delay and Release Time) seem so badly out of whack. Is it just my system somehow?

Again, I'm not laying any blame, or trying to make excuses. But I most certainly did read the manual, several times. I'd be happy to discuss that if you wish. I'm just telling you about the difficulties that I was having and why (in my opinion) I was having trouble understanding how to use the Performance Tool correctly. I'm just glad I can finally get nice-sounding repetitions like I'm supposed to.

@James Boer said:

The other factor that kept me confused was that none of the factory presets for the repetition tool created a convincing repetition effect either (even when played correctly). I'm not sure why all the settings (for First Keyswitch Delay and Release Time) seem so badly out of whack. Is it just my system somehow?

yeah, I have this problem too. i think all the EXS users do. Are you EXS or GIGA?

Anyway, we had some threads a while back where we were discussing how we didn't think anyone at VSL had tested the repetition stuff with the EXS becuase it was incredibly difficult to use and just sounded bad. In the end we only got back the tip that we had to pre-delay the track so that the reps lined up with the beat, and also I came up with a way to chop up the samples and turn them into a 9 part alternation tool preset. CM mentioned at that time that he heard HERB in the background groaning "they are chopping up my library!" in horror. I think this is what Herb is referring to. But honestly, LIKE YOU, I don't give a flying hooteninny. I just want to make music, as fast and as best as it can be done. And this repetition stuff still has not made itself clear to me. I am not sure why it is so confusing to use. I relaly do understand that there IS a correct way to do it, but I just can't get it. Or maybe I did get it, and it just wasn't working or wasn't useful. I don't know.

Anyway, good luck. Let me know your progress.

Cheers, [;)]

Evan Evans

Evan, I happen to use Gigastudio. I guess maybe the same settings were used for both systems, which would tend to make sense, since they're both doing the same work.

Anyhow, Herb need not fear me. The last thing I need to do is waste time is chopping up his library. I'd rather spent time making music with it.

Oh, and BTW, now that I know how to work the silly thing (Repetition Tool), I'm having a lot of fun with it. Like a kid with a new toy. Although, I'm afraid a shinier toy just showed up in a wooden box with 'Tascam GS3' burned into it. A decidedly low-tech way to package a very high-tech product. [[;)]]

Ah ok. Cool.

Evan Evans

Well, I have to support Herb and the library on this... The repetition samples are one of the outstanding features of this library. And yes, the idea is to always be at the recorded tempo or faster, so that the next note of the repetition sample doesn't arrive before you're done with the current one. To be honest, I've been using my own software (MaxMSP) to make use of the perf-rep samples, so I don't use the Performance Tool for perf-rep instruments...
I disagree that the extra repetitions are redundant, since the whole point of the system is to be dealing with an ORIGINAL PERFORMANCE, whenever possible. It seems to me that's the general philosophy of the VSL: to sample performances, not just pitches. And then, after sampling the performance, to find a way of playing it that allows as much flexibility as possible, while maintaining the integrity of the original performance. This is always a compromise, but with the new technologies that are coming "down the pipe", the compromise is getting to be less and less significant. I sincerely believe that it will be possible, in the not-too-distant future, to extract both the essence of a performance AND the sample data, and use them for the creation of new musical material. There are already forms of resynthesis that are basically ready to perform this feat. But the essential thing here is that it will be the PERFORMANCE that is captured, not just the spectral data.
When I read a while ago, Evan, that you never used sampled trills, but used the perf-leg instruments instead, I knew that you had a philosophy which was basically incompatible with the VSL. Taking trills as an example, and speaking in strict terms of "sample signature", it makes less sense to use the perf-leg instruments for trills, than using the trill samples, since the ear (brain, actually) will more easily identify the repetition of an identical sample at shorter temporal intervals (i.e., a trill using perf-leg instruments), than it will the repetition of a longer sample at larger temporal intervals (i.e., the use of the same trill sample at two different points in a piece). (Obviously, in the case of two identical, consecutive trill samples, the ear will easily spot the repetition.) I don't have any problem with your philosophy on this, and I do appreciate your ethic of using only single notes (not performed samples, like trills and runs), but it does kind of place you at odds with the VSL, generally speaking.
My feeling (and I think this is the "vision" of the VSL) is that the ideal situation would be to "sample" an absolutely exhaustive repertoire of complete pieces for each instrument, then extract the sample data from these recordings. In this way, each sample is not an isolated demonstration of a single note, devoid of a performed context, but is rather a "slice" of the musical expression of the performer, taken in the context of a complete musical work.

I hope that made at least a _little_ sense! [*-)]

J.

jbm, your comments make sense, but ... add pitch fluctuations to the already enormous collection of requirements to be able to use 'performances' rather than 'samples'. When the vsl solo strings rise in pitch on most every note, (legato, up section) that works well often, but it really messes up some passages where the REAL performance should have sustained pitch or moved down slightly for tuning.

The whole notion of 'performances' is dangerous imo because it leads to conventional music based on what the recording staff and performing artists think to be most typical. VSL is teetering on that edge already, again, imho and only based on my limited experience with Horizons solo strings.

gugliel,

I hear you... but I think the main problem is size and space, and I think VSL are working on ways around that problem. What I mean is that to sample different styles of performance is no big deal, but it requires more samples -- by nature of the perf-leg instrument process, *alot* more!

I think there are a few methods of pulling more from less that are coming up in the not-too-distant future -- these are methods of sample resynthesis-by-analysis that allow quite a bit of flexiblity in the application of time-stretches and pitch fluctuations to otherwise "concrete" performances.

But your point is good about conventional performances...