Comparison of Samples to Real Orchestra - Can you tell what is real?

[Garritan]

A few months back, Dan did an A/B comparison (http://showthread.php?t=34603) of a sample mockup and a recording of a real orchestra playing a Clara Schumman Concerto and also the Second Movement of Dvoraks New World Symphony. This time he has completed the First movement of the Dvorak (Symphony No. 9 in E minor) and has again done a spectacular job.

Here is Dan's version of Dvorak's New Work Symphony alternating between both real orchestra and samples:

http://www.garritan.com/mp3/Dvorakcompareandsave192kbps.MP3

See if you can tell where the samples are playing and where the real orchestra is playing.

If you can hear the differences, list the spots where you think the real orchestra is playing and where it is samples.

This is something fun and I encourage others to do what Dan did and to emulate the masters. Many students in the colleges and universities often take an orchestral piece and do mock-ups of the classics. This is a great way to get into the mind of a composer/conductor. Also going from MIDI-mockup to real orchestra, as we did in the GPO Competition, is also helpful in learning a great deal about orchestration.

If you can figure out which segments are the real orchestra and which are samples, let us know. The last time some posted that the differences were obvious (after the identity of real/samples were posted). This time let's see if you can the difference without the times being posted.

Thanks again Dan for showing us what is possible. As with learning real instruments, there is a great deal of skill and expertise that your have acquired.

Comments and suggestions are welcome as to how to improve the orchestral rendition.

Bravo Dan!

Gary Garritan

[EDIT: Here is the narrated version with the indications of real and GPO: Here is a narrated version with the times:http://www.garritan.com/mp3/Dvorakc...n-narration.mp3 (http://www.garritan.com/mp3/Dvorakcomparison-narration.mp3)
Thanks to those who took part in this exercise!]

[SeanHannifin]

I could make some guesses, but to be quite honest, this is so well done, I'd rather just sit back and listen to the great piece! :D As far as I'm concerned, Dan is the master!!


This is something fun and I encourage others to do what Dan did and to emulate the masters. Many students in the colleges and universities often take an orchestral piece and do a mock-up of the classics. This is a great way to get into the mind of a composer/conductor.
Very true! And with GPO, you can easily hear each instrument or section separately, an incredible education technique. I will definitely have to start doing some myself!

You always amaze me, Dan!

[Christopher_Reeves]

The distinguishing factor for me was that the strings were bright in GPO

[falcon1]

Gary, yes it could be heard - but very good effort from Dan Kury! Don't want to ruin the fun for the others so no hints from me at this point. :D

[RickD]

I'm out of town listening on 18 dollar headphones with my laptop, and I can't detect any change between GPO and the Sym, though there were a few places where the attacks on strings sounded like they could have been GPO.

What can I say, another stunning job by the maestro!

Rick

[AVBunyan]

See if you can figure out which segments are the real orchestra and which are samples. Dan can perhaps further tweak and refine his orchestral mockup based on your suggestions.

Bravo Dan!
I'm the "low man on the totem pole" here - I cannot hear a difference so this tells me two things:

1. GPO is awesome- the possilities are lunlimited!
2. Dan and others like him are amazing!

I'm inspired! http://www.northernsounds.com/forum/images/smilies/smile.gif

God bless.

[Garritan]

Any guesses?

[Markleford]

Wouldn't dare to guess!

It's fantasitic to be able to do such a convinving mock-up in itself.

It's even more amazing to have it match an existing orchestra.

- m

[DPDAN]

Hi everyone! :)
thank you for the compliments... Christopher, where are the strings bright? ;)
come on Falc, spit it out... :) take a chance, also for those who would consider cheating by using the link from my other topic and lining them up in your DAW, shame on you, you shouldn't have to do that :p

Make some guesses, nobody will be exactly correct. I messed up when I did it. I will tell you that sometimes the crossfades take place at rests, while other times, the splice is in a spot that has no relevance to time.

Have some fun with this, we all could use a little cheering up and some carefree fun in our lives these days.

Thank you for posting this Gary. :)


Dan

[Markleford]

Okay, I'll take one stab. Listening through headphones this time...

4:52 there's series of repeating trumpet notes that are too precise to me, including a little stereo glitch at 4:55. That would lead me to believe GPO.

- m

[Rhap2]

Dan:

As always, a tremendous bit of artistry in mixing for this mock-up. I could guess, but I am at a loss to back up what I think are "real" or "GPO."

You are the best...........

Jack

[joaz]

I am only able to listen through cans on the laptop @ the moment, but my considered opinion is that it is a trick question, and it is all Live Orchestra. :D :) :D
Seriously though, you have raised the bar so high, I am gonna have to get out a spectrograph to check, coz I don't trust my ears anymore.
OK hands up everyone who thinks sample libraries are overated, and we should all go back to synths and soundfonts????? ;)
regards

[fastlane]

I'd guess that only beginning is an orchestra and the rest is GPO. It just sounds more analog to me than the rest of the movement.

[nofish]

It sounds like it switches to an orchestra around 1:15 then back to GPO at about 1:45

[Styxx]

There's just no comparison. I've compared the comparison and have come to the conclusion to compare the comparison there is just no comparison.:D

[Richard N.]

Is that the conclusion of your conclusive comparison conclusion?

[Melody]

I can't tell the difference :confused:

[Theodor]

I know the real parts but i don't wanna spoil the game :- )

[arwa02]

I'll make my guesses, though I haven't used GPO a lot so I don't know what to look for (*cough*).

Hmm, I'd say most of it is sequenced, but very nicely. :) Especially the strings.

My guess is that it's a real orchestra at:
1.05 to 1.44 (but I'm also a bit suspicious about the timpanis just before that), 2.01 to 2.07 and 6.05 to 6.33.

Some parts weren't very audible, so I couldn't really tell, but I'd say the rest is sequenced. Anyway, I don't find it obvious at all what parts are real/fake. Great mockup!

/Arne

[janila]

These comparisons are great fun even if it is somewhat clear that no orchestral library can really pull these things off (even if DPDAN might ;) ). There seems to be a great deal of short GPO clips but most of them only last for a couple of seconds. The choice of reverb in the GPO parts is really nice but most of the GPO parts are recognizable. Still there are some bars where it is somewhat difficult to say whether a real orchestra or GPO is playing and that is quite an accomplishment. It also feels that in some parts GPO is playing on top of the real orchestra but that isn't so, right?

[gugliel]

I'd say almost all is sequenced. The section beginning at 0:29 or so, and through around 1:15, sounded real, and beginning at 2:50 too. Stopped listening at 4 something because the equal temperament began to annoy me. And what orchestra, anyway? Not a pro orchestra at any point that I heard.

So for a midi mockup it's great, very well done, but we wouldn't love music if this were all we had. And this is the first movement, isn't it?

[Sil]

The parts I thought were real:

1:24-1:38
1:54-2:13
2:49-3:12 (strings real, winds not?)
3:31-3:47
4:17-4:27
5:00-5:11
5:23-5:31
6:10-6:24
7:21-7:31
8:09-8:21
8:48-8:51 (nice try!)

It'd be a lot harder if I wasn't so familiar with the piece already. The most realistic part is the violins when playing quickly, but not so loudly, and in a particular mid-range. The cellos were not nearly as ambiguous as I'd hoped (especially at 7:01.) The low brass wasn't "piercing" enough, I can't really describe it. The horns at 6:52 sounded a bit silly.

Are these samples far-miced, or did they have to be placed using a ton of panning and volume control? Much of the time I can just tell by the way the instrument is panned (like the horns and winds) that it has no real orientation, and exists only in the computer world. It's getting closer, though!

[Rach3]

Didn't have time for the whole piece. Just the first few mins.

0 - :15 real
:29 - :48 real
1:24 - 1:47 real
1:53 - real
3:16 - 3:30 very fake

By the way, it will be a bit easier for me if I have GPO and the score at hand.

[Garritan]

BTW, I'm anxious to see Gary reveal exactly which sections are sampled, and which are real. I'm half expecting some kind of trick (i.e. it's ALL GPO), perhaps with the addition of some new products of his which aren't on the market yet.Lee BlaskeLee,
I stated the demo contained "both real orchestra and samples". No "trick" (that's not my style). The sequenced portions are all GPO (with the latest available updates) and not a new upcoming product.

[Garritan]

Yes - an interesting exercise, but a bit off-point.The point was to be an exercise.


The whole "is it real or is it sampled?" question is actually quite banal - the only quesiton that we should be concerned with is "is it good music?" To my knowledge this exercise has hardly been done before. And I am unaware of any other comparisons between samples and a real orchestra. Often A/B comparisons are made between different sample libraries, but why aren't there comparisons to the real thing? Comparing different libraries is comparing expectations of what an orchestra 'should" sound like. Since orchestral realism is often a goal for many sample enthusiasts, why compare different libraries rather than to a real orchestra? This seems to be the best way to acheive orchestral realism. I hope this encourages others with various libraries to do this type of exercise.

As far as the selection, the Dvorak New World is great music to many - a classic!



Bear in mind, too, that we can always write for the strengths of the samples (anybody else noticed this problem?). Not a problem at all. The Dvorak was realized because there was permission to use the recording - not because it would put the library in a better light or allow us to "write to the library". It is a professional orchestra (and we are not at liberty to disclose which orchestra it is).

Thank you for your comments rgames. Do you have a best guess as to which parts are real and which are sampled?

Gary Garritan

[Journeyman]

rgames,

The whole "is it real or is it sampled?" question is actually quite banal - the only quesiton that we should be concerned with is "is it good music?" Actually, you're missing the entire point. The point is NOT "is it good music". The point is that it's impressive that an under-$300 library can come as close as it does. _That_ is the point. Yes, there are other more expensive libraries that will come closer. Yes, there are parts of the clip where it's more obvious than others, but again, that's not the point. The point is that GPO comes close enough to have to listen closely to tell the difference.

To try to change the subject to suit your purposes is disengenuous at best.

[Garritan]

Rach3, Sil, gugliel, nofish and others -I admire your wiilingness to go on the limb and give it your best guess. Thanks for participating.

This is hard to do because we are often guided by impressions as to what "should" sound good or real. There are expectations that go into the mix.

Jascha Heifetz, the famous virtuoso violinist, would often do A/B experiments. Most people will assume that only expensive violins (like Stradivari or Guarneri) "sound really good" and that other violins will have inferior sound. But as Jascha Heifetz proved in many experiments, nobody, not even the critics, could tell whether he was playing his Guarnerius or a modern violin; then, if he announced which violin he was playing, the critics would hear what they expected to hear. So, when he would announce he was playing a copy and go ahead and play the Guarnerius, the critics would complain it didn't sound good. Or he would announce the Guarnerius and play the copy and the critics would rhapsodize over the tone. But the point is, Heifetz could tell. Sure, a Stradivarius or a Guarnerius sounds good, but mainly it is much easier to play, especially if you're Heifetz (or DPDan). Much has to do with the skill of the musician rather than the library.

We'll get some other guesses and we'll post the times soon.

Gary Garritan

[arwa02]

I must add that there's a big difference in mockin up a famous classical piece, and do this comparision, and to do the same with an original composition. First, people who've heard the piece before (performed by an orchestra) will be able to hear the difference much, much easier. On the other hand, you have a recording to use as template for your mockup. You can use the same tempo, dynamics, etc. You can imitate the articulations and emotions expressed by individual musicians on their specific instrument. The samples sound nice, but foremost this is state of the art work, by someone with a great ear for music.

When you're making an original composition, it's another game. Now you have to show your own expressiveness on the instruments. Even if you're an accomplished pianist, it's very likely you wouldn't know how to perform a viola line like a professional player would. Or a clarinet solo. Probably, you wouldn't even be able to sequence (manually) a piano part with the same expressiveness as your own live playing. You create a flute solo, and you think it sounds amazing. Then a woodwind player listens to it, and tells you it sounds terrible.

This is sort of a dilemma, and it's rare to hear an original composition that sound as realistic as this (unless it's very percussion oriented, or uses a very dense texture from the beginning to the end). Still, I'd say most people who buy a library do it to create original compositions. This mockup actually shows how far we are from replacing the real orchestra with a computer. The pieces that imitate a live recording of a real orchestra still outperforms original compositions, by far.

There's a lot of work on developing samples and libraries, but the real bottleneck today is that composers really want true "instrumentalists" to render their lines with the same expressiveness as a real player would. The computer needs to do a lot more of the job! :) so that people can focus more on the creative part!

/Arne

[DPDAN]

It also feels that in some parts GPO is playing on top of the real orchestra but that isn't so, right?
Hi Janila, I would never do that,
that would spoil the fun, and besides it would be very misleading and dishonest. Thanks everyone for the fun listening.

Sometime soon, Gary will post the exact times.
thanks
dp

[DPDAN]

The point is that it's impressive that an under-$300 library can come as close as it does.

Amen! http://www.northernsounds.com/forum/images/smilies/smile.gif

[DPDAN]

What an interesting observation Lee. Also when Mr Leslie took his rotating speaker design to the Hammond Clock Co. that had just started making organs, they laughed him all the way out the door. Now, many years later you would be hard pressed to find a Hammond B3 or C3 without a Leslie, it is half the sound. I just learned at the NAMM show at the Hammond booth that Hammond now owns Leslie, go figure. At least they swallowed their pride, but rest assured, someone that used to work for Hammond is rolling over in his/her grave. ;)

Strange how times do change, and yet some things stay the same.

A Hammond is no match for a pipe organ, and vice-versa. They are both, (as you say) unique in their own ways. I would not want to be without either.

Dan :)

[Theodor]

Hi DPDAN,

I forgot to congrat you on your amazing artistry and mockup. You've done a wonderful job there! The expression, dynamics and flow are excellent :- )

[DPDAN]

Theodor and Lee and all the others, thanks alot!

It's all for fun really. I got into midi in April of 2004 and had Digital Performer for about 6-7 months before that. I decided at that point that I should probably learn this midi stuff. So I purchased GPO to help me learn the midi side of DP. No music training except I have been a recording/sound engineer for the last 34 years of my life (49) so that does help to know what instruments should sound like in context, well at least to me. :eek: I am just having too much fun with GPO.
Thanks for your encouraging words.

Dan

[synergy543]

....No music training except I have been a recording/sound engineer for the last 34 years of my life (49) so that does help to know what instruments should sound like in context, well at least to me. :eek:....Hi Dan, I always enjoy listening to your work. I think you under-rate the value of your experience by a "tiny bit". ;)

[DPDAN]

Hi Dan, I always enjoy listening to your work. I think you under-rate the value of your experience by a "tiny bit". ;)
I probably do, thank you for your kind thoughts. To read something like....
"I always enjoy listening to your work" makes me very humbled.

Thank you !!!!!
Dan :)

[synergy543]

I probably do, thank you for your kind thoughts. To read something like....
"I always enjoy listening to your work" makes me very humbled.

Thank you !!!!!
Dan :)Thank you Dan!

I can appreciate how much work it must've been to sync to an existing performance - not only in tempo but in numerous other ways (dynamics, tonal balance, stereo imaging, etc.). Quite impressive. I don't think I've heard anyone else attempt this. It must be one of the most enlightening ways to learn to I'll bet. I assume you worked from stereo performance and didn't have access to soloed tracks? (wouldn't that be great - Orchestral Karaoke - a great idea sample developers ;) )

In the end, I do hope you post both the original and your realization in full so that we might be able to study them.

[DPDAN]

Thank you synergy543,
I did have the real orchestra recording imported into a stereo audio track in Digital Performer (MAC) so that I could sync to it. Of course it all had to be done by ear and alot of editing. Once I had the audio tracks rendered from the PC (GPO), to the MAC, I simply mixed according to my own taste. I really honestly did not try to duplicate the exact stereo placement of the various instruments to accurately copy the real orchestra recording. This is evident in listening closely, many of the woodwinds are not in the same place, second violins, etc. that makes determining which is which reasonably easy. I did have my GPO piece in it's entirety posted in another topic on the Garritan portion of this forum. I will re-activate that link later.

Thanks again for your participation.

DPDan

[Glenn]

Well, I'll hazard some guesses. I thought it jumped back and forth a lot between real and GPO...but maybe I was just fooled!

I thought the piece began with the real orchestra.

Here are some points I noted:

00:20 GPO comes in
00:50 Real
01:08 GPO
01:18 real

01:29 - 1:40 GPO

01:56 real

02:20 GPO but 02:29 back to real

02:49 GPO

02:59 real


I stopped logging at this point. But that's how I heard it.

Was I close?

--- Glenn

[Bruce A. Richardson]

The distinguishing factor for me was that the strings were bright in GPO

What on earth is a composer/arranger doing living in Kilgore????

Welcome (grew up in Longview).

Bruce

[synergy543]

OK you guys have had long enough. Despite all of the Golden Ears here nobody is even close so here it is:

0m0s - Real

0m15s - GPO

0m41s - Real

0m50s - GPO

1m26s - Real

2m14s - GPO

2m25s - Real

2m37s - GPO

2m45s - Real

3m48s - GPO

4m17s - Real

4m28s - GPO

5m0s - Real

5m12s - GPO

5m24s - Real

5m44s - GPO

6m10s - Real

6m26s - GPO

7m12s - Real

7m32s - GPO

8m10s - Real

8m22s - GPO

8m50s - Real

8m53 - END - GPO


OK Gary so what do I win? An upgrade to GPO Advanced? :)

And DPDan - I only know because of my 30 years engineering experience too. ;)

[Garritan]

What on earth is a composer/arranger doing living in Kilgore????
Welcome (grew up in Longview).
Bruce
And a very fine composer/arranger indeed.

Craig is a student studying music. He was one of the top winners of the GPO Orchestration Competition and had his work performed live at the competition concert. You can see a video or hear an mp3 of Craig's 'Symphony No. 1 in G minor: Movement IV (Presto)' here: http://www.garritan.com/competition.html

If you would like to meet Craig, he'll be at TMEA this upcoming year.

Gary Garritan

[Brian2112]

If you would like to meet Craig, he'll be at TMEA this upcoming year.

Gary Garritan

Count me in! it would be an honor to meet him...and he lives like 30 min away from me!

...2112

[synergy543]

I was certain I was almost spot on (maybe missed one or two but I thought my call was pretty accurate) .....yet no response so I guess I wasn't even close? :o

How embarassing is that? :(

Time to get my hearing aid checked again I guess. :cool:

[DPDAN]

synergy543,
some of your times them were spot on, others not, and many times the ones that were off were in favor of GPO being thought of as real.

It's all fun, I can tell you listened critically though. :)

Dan

[synergy543]

Thanks Dan,

I'll be very curious to see the official timings when you post them.

Greg

[grumpkin]

[QUOTE=Garritan]Rach3, Sil, gugliel, nofish and others -I admire your wiilingness to go on the limb and give it your best guess. Thanks for participating.

This is hard to do because we are often guided by impressions as to what "should" sound good or real. There are expectations that go into the mix.

having worked in film music for, ouch, about 34 years i was impressed. but speaking of the mix, trying to compare two mp3's? i would really love to download 2 24 bit files and have a listen in the studio.

anyway, nice work.

[doob]

hi guys,
just stumbled across this site from a search on orchestral samplers (i'm relatively new to the stuff...i don't own garritan and i've never heard it...) and thought this would be a pretty fun thing to try while taking a break from my composition assignments.

0-real
0:17-GPO
0:28-real
0:50-GPO
1:14-real
1:22-GPO
1:29-real
1:40-GPO
1:55-real
2:17-GPO
2:27-real
2:40-GPO
2:50-real
3:14-GPO
3:31-real
3:48-GPO
4:17-real
4:30-GPO
5:00-real
5:15-GPO
5:24-real
5:36-GPO
6:12-real
6:26-GPO
6:53-real
7:01-GPO
7:11-real
7:30-GPO
8:00-real
8:18-GPO
8:48-real
8:52-GPO

i got the times from just a once-through listening and they are probably about +- 2/3 secs off (or a lot more if i was wrong!) because it took awhile to type out the numbers. and if i'm not mistaken i think i heard the synthesized orch and real orch panned to L and R at around 8:00. or maybe my ears were fooling me ;)

anyways, it looks like a great resource site and i'll be sure to be checking back here more often. (and looking forward to seeing the real times too!!)

[synergy543]

Hey Doob,

A lot of your timings are suspiciously close to mine.;) Refined my list?
Or maybe we have the same shaped ears? :D

I was pretty sure I got at lot of those right (at least an A-) but maybe not....we'll have to wait and see.

One pass...hmmm.....pretty good. :cool:

[doob]

hi synergy543,
yep, saw your post but decided not to read it until i'd done my own.
i guess some of the bits are dead giveaways anyway. like the synth lacking the warmth, inaccuracy, and tuning clashes of real orchestras, not to mention sound artifacts (lack thereof), and the solo instruments being waaaay too clear for a real orch ;)

[synergy543]

Well Doob, I own VSL, EWQLSO, GPO, DC_Studios and Miroslav and there are wonderful aspects as well as flaws in all of these libraries. However, I do have great admiration DPDan's impressive comparison which is quite an exercise and must've been a lot of work. I'm sure it was an incredible learning experience too. Well if you know of any other comparisons you would recommend, I would be most interested to hear them too.

I'm not sure that if you used all the libraries above you could necessarily come any closer to an existing performance for there are tonal differences between intruments and interpretive differences among players, conductors and orchestras. I think what we can do (and even moreso in the future as tools advance) is go beyond mockup and emulation with these tools to create interpretive and expressive performances that stand strongly on their own. This is a wonderful time for composers - the orchestra has been democratized.

[vegas]

are the times going to be revealed soon?:confused:

[DPDAN]

Yes they are :)

Thanks everyone. I think I mentioned this before, that I did not try to duplicate the recorded orchestra's exact sound. I only made a midi/GPO rendition and made them match time. Because of this, it did make it easier to detect the difference by simply listening in a quiet, non confusing environment. The mere idea that we can get as close as we can with these libraries is a very cool thing.
I for one am able to make some music for the first time in my life.
It has been a fun ride.

Dan

[Garritan]

Kudos to those who were brave enough to guess and some got remarkably close.

Here is a narrated version with the times:
http://www.garritan.com/mp3/Dvorakcomparison-narration.mp3
And here are the times for the original file (which is linked in the first post of this thread:):

0:00 Real
:15 GPO
:40 Real
:49 GPO
1:24 Real
1:38 GPO
1:53 Real
2:13 GPO
2:24 Real
2:36 GPO
2:45 Real
3:11 GPO
3:30 Real
3:47 GPO
4:16 Real
4:27 GPO
4:59 Real
5:10 GPO
5:23 Real
5:30 GPO
6:09 Real
6:24 GPO
7:10 Real
7:30 GPO
8:08 Real
8:20 GPO
8:48 Real
8:53 GPO
Thanks to Dan for doing this. It must have been very time consuming to do this comparison and takes a remarkable skill to do so. What is amzing is that one person can do this - Dan wore the hats of conductor, engineer and each and every player of the orchestra.

If you listen closely the differences become more apparent. Although this is a fun exercise and samples can be an excellent learning tool, it is not a replacement for the real thing. Orchestras are made up of people who devote their lives to perfecting their instrument. This will not be replaced by static samples anytime soon.

Thanks for listening.

Gary Garritan

[DPDAN]

Thank you everyone again for listening.

I would like to add a note to Gary's post,,,,

The times that are posted are for the original posted link without the narration.

Use this link (http://www.mydocsonline.com/pub/DPDAN/Dvorakcomparisondemo192kbps.MP3) that coincides with the crossfade times.
This link is the piece without the narration. Use it to listen to if you want to simultaneously look at your guesses.

Good job by so many, thanks again everyone!
And thank you Sir Gary http://www.northernsounds.com/forum/images/smilies/smile.gif

I would also like to take this opportunity to thank all the folks at MOTU for the incredible editing and mixing features in Digital Performer.

dpDan

[hollowsun]

Hats off to Garritan and especially Dan... very impressive.

I did have my own thoughts about the original MP3 (some right, some not) but I wasn't gonna hazard a guess in public ;)

Funnily enough, now that the 'official' times have been published (and listening to the voice-over version), you can hear it more clearly but NOT because of any deficiencies in the playing/arrangement - simply because the GPO's first violins are just a tad 'wider' in the stereo image than the original... but only just - it's subtle.


I would also like to take this opportunity to thank all the folks at MOTU for the incredible editing and mixing features in Digital Performer.
Indeed - with respect, the edits are almost as impressive as the GPO/Dan 'arrangement'.

Nice job to all involved - I am with Gary that samples can never be a replacement for 'the real thing' (especially live - staring at a sequenced laptop in a concert hall is no substitute for watching a bunch of musos on full tilt ;) ) but it's a testament to how good a quality lib like this can sound in the right hands (it would probably sound a right old mess in mine!!!!).

I just wish I had my old 'trons so that I could do a similar 'real/sampled' exercise ;)

[DPDAN]

So kind of you Steve, thank you so much!


Dan

[hollowsun]

So kind of you Steve, thank you so much!
Hey... come on - credit where it is due!

[Hans Adamson]

First I was impressed with DPDan's magic touch when mixing in different segments of sampled and real orchestra. 34 years of recording engineering experience really shows off in an exercise like this...

Then I was also equally impressed with Synergy543's accurate pinpointing of real and sampled segments!

So, I decided to compare how well the other combatants did compared to him, and found that everyone who made a serious attempt had pinpointed the real and sampled segments with very high accuracy. Bravo!!
Here are the results:

DPDAN Synergy doob Sil Rach3 rgames

0:00 Real 0m0s - Real 0-real 0 -real
:15 GPO 0m15s - GPO0:17-GPO 16-GPO
:40 Real 0m41s - Real0:28-real :29-real 0:30-GPO
:49 GPO 0m50s - GPO0:50-GPO :49-GPO
1:14-real
1:22-GPO
1:24 Real1m26s - Real 1:29-real 1:24-real1:24-real 1:24-real
1:38 GPO1:40-GPO 1:39 GPO 1:48-GPO 1:48-GPO
1:53 Real1:55-real 1:54-real 1:53-real 1:59-real
2:13 GPO 2m14s - GPO2:17-GPO 2:14-GPO 2:14-GPO
2:24 Real 2m25s - Real2:27-real 3:31-real
2:36 GPO 2m37s - GPO 2:40-GPO
2:45 Real 2m45s - Real2:50-real 2:49-real
3:11 GPO3:14-GPO 3:13-GPO3:16-GPO
3:30 Real3:31-real 3:31-real 3:30-real
3:47 GPO 3m48s - GPO3:48-GPO 3:48-GPO 3:52-GPO
4:16 Real 4m17s - Real4:17-real 4:17-real 4:17-real
4:27 GPO 4m28s - GPO4:30-GPO 4:28-GPO 4:29-GPO
4:59 Real 5m0s - Real5:00-real 5:00-real 5:00-real
5:10 GPO 5m12s - GPO5:15-GPO 5:11-GPO 5:15-GPO
5:23 Real 5m24s - Real 5:24-real 5:23-real 5:23-real
5:30 GPO5m44s - GPO5:36-GPO 5:32-GPO 5:35-GPO
6:09 Real 6m10s - Real6:12-real 6:10-real 6:02-real
6:24 GPO 6m26s - GPO6:26-GPO 6:25-GPO 6:44-GPO
6:53-real
7:01-GPO
7:10 Real 7m12s - Real7:11-real 7:21-real 7:02-real
7:30 GPO 7m32s - GPO 7:30-GPO 7:31-GPO 7:36-GPO
8:08 Real 8m10s - Real 8:00-real 8:09-real 8:02-real
8:20 GPO 8m22s - GPO8:18-GPO 8:21-GPO 8:25-GPO
8:48 Real 8m50s - Real8:48-real 8:48-real
8:53 GPO 8m53 - GPO 8:52-GPO 8:52-GPO

[DPDAN]

Thank you Steve :)

Holy cow Hans, that took some time to organize, thanks alot for doing that.

Dan

[Garritan]

Hans,

Thank you for taking the time and the interest to attempt to sort this out.

Actually it seems that it was doob who was closest. Congratulations doob! Contact me by email and you'll get a library.

There are 28 different changes. Some of what is identified in Hans' chart was off by as much as 22 seconds. If we look at the results within 10 seconds of the actual mark, the results are as follows:

doob's accuracy rate was 93%. doob identified all but 1 within 10 seconds of the actual times, however, he did misidentify a couple of transition points that weren't there.

Synergy missed four transitions and was more than 10 seconds on one of them and was estimated at 89% accurate

sil had about a 68% accurancy

Rach3 didn't do all of them but was accurate at what he identified and had an overall 29% accuracy

rgames got 17 of them or 61% accuacy.

It we lower the threshold where a miss is 5 or more seconds then the accuracy rate would be less.

An outstanding job was done by all who took part. We'll be working on another exercise and perhaps Hans can help us with the next one.

Gary Garritan

[Hans Adamson]

Thanks Gary,

You are welcome. I enjoyed it - it was fun. Don't you think the guys were amazing?

I think your statistics may be a little off though. I continued today's brain excercise and made a chart of the two up-runners scores! http://www.northernsounds.com/forum/images/smilies/smile.gif

Synergy543 missed two entries, and doob made up two entries that didn't exist. That should equal out each other. The rest of the entries can best be measured in seconds OFF. Synergy543 was totally 41sec off the mark and doob was totally 89sec off the mark.

The chart below shows how many seconds they were off each entry:

DPDan Synergy543 doob

0:0 0 0

0:15 0 2

0:40 1 12

0:49 1 1

1:24 2 5

1:38 - 2

1:53 - 2

2:13 1 4

2:24 1 3

2:36 1 4

2:45 0 5

3:11 - 3

3:30 - 1

3:47 1 1

4:16 1 1

4:27 1 3

4:59 1 1

5:10 2 5

5:23 1 1

5:30 14 6

6:09 1 3

6:24 2 2

7:10 2 1

7:30 2 0

8:08 2 8

8:20 2 2

8:48 2 0

8:53 0 1
Total: 41s 89s

Maybe you should just give them both a lib! http://www.northernsounds.com/forum/images/smilies/wink.gif Or give them all one lib each - they were all really close!

Best,
Hans

[Garritan]

Hans,

Crossfades can account for time so an aggregate of the deviation was not the test. If someone was off within a second or two the crossfade would account for that. The statistics were not off since missing entries and inserting entries that didn't exist were not considered to equal each other.

I appreciate all the efforts you made to compare over 100 separate time entries and go through the trouble to color code each one and then add up all the seconds. That was very magnanimous (and tedious) of a fellow sample developer to go through all the trouble. Nevertheless, this was not a contest but an exercise. The offer of a library was made as a goodwill gesture after the fact.

Both will get a library. Congratulations to Synergy and doob for getting close as they did. Perhaps you could offer a piano library too . :D

I welcome comments and suggestions on how to make the mock-up even better the next time.

Gary Garritan

[Hans Adamson]

That's really nice of you to change the post and give them a price each!

I appreciate it!!

Hans

PS Sorry I didn't see that you also added: "go through the trouble to color code each one" - Hehe, no trouble. It's just the old teacher in me...http://www.northernsounds.com/forum/images/smilies/smile.gif Don't make any more changes now! - I'm going to bed http://www.northernsounds.com/forum/images/smilies/wink.gif
Yours truly

[Garritan]

Count me in! it would be an honor to meet him...and he lives like 30 min away from me!
...2112
Brian, it will be great to see you and Craig there. I'll be teaching a clinic at the TMEA this coming year. Come with no-doze and tomatos! http://www.northernsounds.com/forum/images/smilies/ninja.gif

[bbybeaniej]

This is totally unfair!! The GPO-only version was posted several weeks ago. All somebody had to do was compare the GPO file with the the a/b comparision and that would have clearly shown which parts were which. If I would have known this was a contest maybe I could have won a free library too!

[kid-surf]

I only listened to the clip and read the first page, but this is my thought. (so i don't know if this has already been said?)


The difference was pretty obvious in spots. I'll tell you why. For one, the real session sounds more compressed (a compressed quality -- not pumping and such, like tape compression), the fake parts sound wider, and well, a lot more disconnected (more space in the mix). Then there is the attack of the strings, on the real one it's like a "tsssszing" every time it bows, with the sampled it's like a "mvvvuummpphh". I don't know wow else to describe it, just a more blunted sound with samples on the attack.

There were a few parts where I wasn't sure, and then some that I was positive which was which.....


Interesting to hear it this way though............

[DPDAN]

Kid surf, there was absolutely no compression of any kind on either the real or GPO version.

Not sure what you were hearing, but no compression.
Thanks for your comments.

Dan

[Rach3]

The fact is: samples can sometimes fool the best ears. There's no doubt about it. After all, samples are recordings of real performances.

I think this kind of exercise is to the advantage of the mock-up arranger, simply because the arranger can decide on which part is sampled, which part is real. The most difficult part for the arranger is to match the mix, and dan has done a marvelous job in this area.

For example, after 0:15, the GPO strings tremolos come in. A sampled tremolo string pad can sound just like the real thing. However, in this case, the difference in ambience gave it away.

It is surprising that almost no frequent posters here participate in the real guesses. Maybe they are just too busy...

Again, I wish I had done the whole thing. Then maybe I would be lucky enough to win something from Mr. Garritan!

[kid-surf]

DP --- Well, like a side it was a sense of compression, not actual compression (maybe I wasn't clear in what I was saying). What I'm saying is the samples don't have that sort of natural compression that a recording has (even when the recording isn't "compressed" per say, know what I mean?). That is where they sounded the most different in a spatial way IMO. The other was textural..........


BTW --- I meant tape compression not in a "slamming the meters" sense just in the sense of gluing it together.....


If you are the one who did this, well done. Not trying to take that away from you (or whoever?) just telling you what my ears hear.

Rach3 --- yeah, I myself am too busy to go through the whole thing and type out what I think is what... I would bet many are.

[synergy543]

I just returned from a trip to find the surprising direction this thread has taken.

Hans - thank you very much for your meticulous observations. I hope you're not taking time from your next sampling project as I'm absolutely loving Malmsjo and the Virtual Grand too much to stop you from working on the next one.

Gary - Thank you for your kind offer and making me feel like one of the winners. However, I truly didn't know this was a contest either and just posted for fun. Since I already own GPO and I can't give it away, I would much rather have you give it too someone in need.

DPDan - Hey, I envy you're making the time to do this emulation. Its something that I've wanted to do to - as a learning experience. I think it must be the best way to learn on your own. Well, I don't want to force your hand into sharing your golden nuggets but... if you want to, I'd be most interested to hear what you learned and gained the most from doing this and what you would suggest others focus on when trying to do an emulation such as this.

Gary made an important point in the opening of this thread:

This is something fun and I encourage others to do what Dan did and to emulate the masters. Many students in the colleges and universities often take an orchestral piece and do mock-ups of the classics. This is a great way to get into the mind of a composer/conductor. Also going from MIDI-mlockup to real orchestra, as we did in the GPO Competition, is also helpful in learning a great deal about orchestration.So regardless of the library you use, we can all learn by trying these ideas to improve our own skills and techniques. This truly is a renaissance period in music and the possibilities of what we can do with these new tools is quite amazing as DPDan and others are showing us.

I look forward to further discusssions about the process of what DPDan has done. Has anyone else tried this?

[Sil]

That is where they sounded the most different in a spatial way IMO. The other was textural..........
I know exactly what you mean. I'm not too familiar with how the instruments are recorded for sample libraries, but I'd like to believe that if the players were sitting in an orchestra with the microphone(s) placed where they would naturally be in professional recordings, the sound would turn out to be more realistic. They're probably doing this already, I don't know. It doesn't always sound like they are.

Anyway, maybe the strings and brass need some kind of compression to remove some of the brightness and "ringing", and a narrower spatial range so things don't seem artificially panned. I converted the file to 64 kbps/22050Hz and was surprised at how much more some parts sounded, albeit lower quality overall.

[Hans Adamson]

If I would have known this was a contest maybe I could have won a free library too!
I don't think you necessarily have to give away prizes when making a promotion like this. It is nice to see though, that the contributors get credit! http://www.northernsounds.com/forum/images/smilies/smile.gif

Hans

[synergy543]

This is totally unfair!!.....If I would have known this was a contest maybe I could have won a free library too! Do you not have a sound library? As I said, I already have GPO so if you're in need maybe you can e-mail Gary. Even if you have to buy it, its not too expensive considering what you get and the requirements to run it. And if he gives it to you for free, you can still save up to buy the Advanced Version!

[Garritan]

Thank you for your kind offer and making me feel like one of the winners. However, I truly didn't know this was a contest either and just posted for fun. Since I already own GPO and I can't give it away, I would much rather have you give it too someone in need.
synergy,

This wasn't intended to be a contest. I think Hans may have misunderstood the exercise.

The all-GPO version was posted, as bbybeaniej pointes out, and it would have been easy to compare the waveforms of the two files. If it was a contest, the original file would not have been public.

Nevertheless, the offer of a library was just a goodwill gift to recognize how well doob did (not a prize). Hans pointed out that viewed another way your results could be considered as most accurate. (although Hans wanted to put my library in the hands of more people). So I reconsidered, did the Solomon thing, and offered a library to both doob and you.

It was very generous of you to offer the library to bbybeaniej. If bbybeaniej will contact me I will send a library.

Perhaps in the future we can do a formal contest.

Gary Garritan

[synergy543]

Gary, you are very generous. And it is good of you to host and encourge these mockups as its a great way for others to learn. Musicians in the past have always learned by imitating and copying eachother - its only recently that litigation has come into the equation!

Personally, I'd much rather see us analyze and study together rather than compete with the WWF as I can watch that any night on T.V. But discussing orchestration, and mockup techniques, now that's special!

btw, this would be a good opportunity to promote the numerous tutorials on your site. There are some real gems there.

Thank you!

[Hans Adamson]

Gary,

I have never referred to your promotion as a contest. Neither have I suggested that you give prizes until you handed out one to doob:


Actually it seems that it was doob who was closest. Congratulations doob! Contact me by email and you'll get a library.

Gary Garritan
When members of the forum play along and put some effort into supporting your promotion by setting aside time to do an exercise like this, I don't think it is more than fair that you at least recognize their effort, and note their correct answers. That's just my take on it. No need really for prizes or competitions.

Hans

[vic_france]

... So I reconsidered, did the Solomon thing, and offered a library to both doob and you.

Gary GarritanNo, the "Solomon" thing to do would have been to offer HALF a library to each of them... which places you better than Solomon in my books! :p :)

[DPDAN]

No, the "Solomon" thing to do would have been to offer HALF a library to each of them... which places you better than Solomon in my books! :p :)

LOL!

Dan