Too real, or not too real, that is the question

All the official new product / update announcements from the Samplemodeling team
Site Admin
Posts: 74
Joined: Fri Mar 07, 2008 9:22 am

Re: Too real, or not too real, that is the question

Post by webmaster » Mon Feb 11, 2013 5:00 pm

Dear Friends,

many thanks for the participation in our experiment... :-)

There were no tricks, traps or misleading elements. The 5 versions were played in real time either "out of the box" on the EWI, or on an alto sax using different mouthpieces. No EQ or compression was used. The same background and reverb was applied to each version.
The truth is:

1 real
2 virtual
3 real
4 real
5 virtual

Congratulations to the winners: Pier-V, tobinho and lafauna! Please get in touch with us via e-mail/Contact page. Maddcow was also right, Kobir and Plougot kind of "half-right", since they finally opted for a wrong guess... :-)

So the statistics is:
18 guesses, 4 correct, 2 uncertain, 12 incorrect :-). This means that more than 75 % could not fully distinguish between real and virtual instruments.

Can we ask you what were the clues which allowed you to quickly identify the right versions?

The main purpose of the experiment was to gain some further knowledge about still existing differences between real and virtual saxophones. We want to learn and improve our instruments whenever possible...
Thank you for your collaboration.

Peter & Giorgio

Posts: 205
Joined: Wed Jan 30, 2013 12:20 pm

Re: Too real, or not too real, that is the question

Post by Kobir » Mon Feb 11, 2013 5:31 pm

I am happy to assist you in improving this already amazing virtual instrument. I have never enjoyed playing an instrument as much as SWAM saxes, even my real alto sax was a bit neglected recently.

The main clues that I used:

* Beginning notes start more gentle in real instrument compared to virtual in this recording (check the note the starts after 19 seconds - huge different between real and virtual in all versions). This might not be an issue with the virtual instrument, but with the the EWI playing, since I have been able to get smoother beginning notes with my wind controller.

* Real instrument have smoother transitions between notes in general, especially between the short notes.

* Short notes is a weakness of the virtual instrument and you have many of them in this song. They are not as sharply defined and colorful in the virtual compared to the real.

* Real instrument have more color and dynamics (same note doesn't sound the same when you play it again and again) and the difference in volume between individual notes is much bigger in the real vs. the virtual (especially around 1 minute of the song).

I am really sorry I didn't get to the forum quick enough to guess. I was really hoping to get the Trombone...

Keep up with the amazing groundbreaking work and let me know if you need any other feedback, demos or beta testing support for wind controllers.



Posts: 159
Joined: Sat Mar 24, 2012 10:45 am

Re: Too real, or not too real, that is the question

Post by Plougot » Mon Feb 11, 2013 5:40 pm

I must say that my first choice was the good anwser, but since there was no hope of winning with this combination, I tried something else. If it had been stricly a "guess for free contest", i would have been right ^^.
But not without a lot of hesitations, i must admit.
So, what are the clues ?
For me, the first one is that i "know your sound". Obviously, you have a really clean sound, way cleaner than most of the recordings (which doesn't help blend in). And since each microphone has kind of its own "signature", your sound is recognizable. Which doesn't mean it sounds fake.
the second and fifth example have some dynamics and attacks "flaws", which make it pretty easy to guess. (00.13 or 00.25 for example)
In the forth, they're kinda here too, but not as "sharp".

IMHO, the clues are not so much coming from the instruments themselves, but from everything around them. Blending a dry instrument into a mix leaves clues. Playing an instrument on a windcontroller with a different dynamic response, without feeling the instrument "vibrating", leave clues...
Not having tricks nor traps also leaves clues, because you usually grab some noise of some sort in a recording...
Not to mention we had several versions to compare. Wouldn't as easy on several different pieces.

But I think it's fair to say that your instruments are far better than any other on the market. And i'm pretty sure that i wouldn't recognize them i a normal context.

Posts: 27
Joined: Wed Sep 30, 2009 11:08 am

Re: Too real, or not too real, that is the question

Post by lafauna » Mon Feb 11, 2013 7:38 pm

Hello. I will be short, because of my very limited english.
sorry about that.
Really happy to win! thank you very much Sample Modeling.

My main cue to hear the diferences, is that in virtual ones,
the first harmonics remains the same sound in dinamic
changes, when played soft, I hear a sound, when played
loud I hear that sound with added higher harmonics, so
I hear two sounds mixed, instead of a one more complex sound.

Also hear a more natural vibrato in the real ones.

I'm enjoying your saxes alot, playing with my wind controller, your instruments are the most playable.
Hope my english is understanable.
Kind regards

Posts: 168
Joined: Tue Aug 12, 2008 10:51 pm

Re: Too real, or not too real, that is the question

Post by Phil999 » Mon Feb 11, 2013 8:18 pm

I agree to the other people. There are fine nuances, difficult to describe, that still make a difference between a real instrument and a Samplemodeling instrument.

There is some 'roughness' or 'grittyness' in a real sax, which is quite impossible to translate to a virtual instrument. Of course, the attack and the end of a note is more relevant and discernible, but also the sound itself gives clues.

By the way, I listened to the samples only once, and not even entirely through, and my judgment was right. I just chose differently because there already were three correct answers. Which could have been wrong, I wasn't sure at all. But from a brief listening I could discern real from virtual without problem, obviously, now that we have the results.

Now, how could Samplemodeling instruments improved? That's a difficult question. How do you add 'grittyness' to an already perfect sound sample? There are two ways, as far as I know. First the easier one, adding modulation and/or audio processing like saturation algorithms. Quite common today with analog synth emulations. The second is much more difficult I think: add more samples. The player not only morphs through different layers with the expression CC, he also morphs through different playing styles/sound variations with the portamento CC (or another one). And on top of it, a round-robin fluctuation that the player doesn't influence at all.

It is already there with the growl and flutter parameters, but I don't exactly know how these effects are done. Modulation/post audio processing, or samples? My guess is the first.

If my guess is right, the latter would probably be too complicated for a virtual instrument, even for Samplemodeling. It would quadruple the install file size or more, would involve much more work in sampling and scripting, . . .

I don't know if this is really worth the effort. If you can play the wind controller or the keyboard with expression pedal/mod wheel, the Samplemodeling instruments give you a nearly perfect emulation that most people couldn't discern from a real instrument.

I am very demanding in music topics, but I feel entirely happy with Samplemodeling instruments since the beginning. They are good enough, and still miles ahead of the competition.

Posts: 159
Joined: Sat Mar 24, 2012 10:45 am

Re: Too real, or not too real, that is the question

Post by Plougot » Mon Feb 11, 2013 10:06 pm

Maybe it's just because it's still sort of sample based. Maybe the only way to step up to the next level would be to completely recreate the instruments in 3D, and then calculate the air flow, the lips of the user, all the little subtleties that go with it...
Of course, it's out of reach for now, it would probably, as Phil999 said, not be worth the effort, and it wouldn't even work on our present computers :p.
That's one big difference between image and sound... Image doesn't need realtime...
But hey, one step at a time. As far as I know, you're still the only ones building the future of virtual instrument.
You'll deal with next next gen virtual instruments later^^

Posts: 41
Joined: Wed Jan 23, 2013 11:15 pm

Re: Too real, or not too real, that is the question

Post by kalkrishnan » Tue Feb 12, 2013 5:49 pm

Great contest, and good answers on how the winners did it. Lot of lessons to be learned by all.

One word of caution to Samplemodeling: It is indeed very heartening to note that you are pursuing perfection - you have taken this approach to answer the question "how can I make it more perfect?". However, one consequence may be "too much perfection"! What does it mean? For example, a "perfect" saxophone won't be able to do a smooth portamento/glissando that spans 3 or 4 semitones, and the fact that your model does allow it will make it sound "fake" when used. But, this is what makes THIS instrument unique and fun. In other words, let's make sure that this quest of perfection does not perfectly imitate the limitations of the acoustic instrument. The fun of a digital instrument is that it CAN realistically emulate an acoustic, but can also offer added expressive abilities due to the synthetic nature.

One approach would be to have a mode switch: strict emulation vs. relaxed emulation - the former is for people who want to really fool the audience (and are willing to live with the limitations imposed), and the latter is for people like me who like to experiment with the sounds for alternative uses!

Posts: 7
Joined: Thu Sep 22, 2011 9:04 pm

Re: Too real, or not too real, that is the question

Post by tobinho » Tue Feb 12, 2013 10:31 pm

Very happy to win :)

First I would like to thank the Sample Modeling team for making the most flexible, expressive, playable and in almost every situation the best sounding instruments. I could write a book about how happy I am with your instruments.

Some thoughts on the contest and Sample Modeling instruments in general:

- I identified the real examples from the virtual ones due to the characteristic anechoic chamber sound your instruments possess. Also the real sax behaves more naturally in the room. Maybe you could create a spatialization/reverb plugin like VSL's Mir Pro with directivity profiles for all your instruments?

- The sax in the real examples has these small human, random imperfections when playing. The virtual sax has a more static sound.

- Legato transitions sound better and more natural in the real example.

Generally speaking your instruments sound awesome in most situations, styles and tempos. As to the timbre/style of your instruments I feel we could use a more classical sound for some of the instruments especially for the trumpet and clarinet (nice you added a clarinet with a less jazzy sound). As to the legato of your instruments I sometimes feel a real recorded legato interval transition is preferred over yours.

Thanks for making my favourite virtual instruments and I hope your achieve perfection ;)

Posts: 17
Joined: Wed Sep 24, 2008 12:31 pm

Re: Too real, or not too real, that is the question

Post by MassMover » Thu Feb 21, 2013 9:46 pm

I found it quite obvious that the first note (g) of the real saxophone was permanently out of tune thoughout the whole piece.

As virtual instruments are tuned perfectly, this was the one more clue.

Post Reply

Who is online

Users browsing this forum: No registered users and 1 guest