Imperfect shorts test - Berlin Strings (SINE)

[Henrik B. Jensen]

Warning: This is an imperfect test! But hopefully still useful for a certain purpose (more below)

This is Berlin Strings (SINE) playing some Staccatos OOTB (and please excuse my playing):
View attachment Battlefield 1-inspired - hbj4 - Berlin Strings SINE OOTB.mp3

Now, let’s keep everything the same as above, except we turn off the top dynamic layer, ff.
That sounds like this:
View attachment Battlefield 1-inspired - hbj4 - Berlin Strings SINE OOTB except ff-layer off.mp3

Now, let’s keep everything the same as in the OOTB example above (the first example), except we turn off both the top dynamic layer, ff, and the one below that, f.

Remaining dynamic layers are p and mf.

That sounds like this:
View attachment Battlefield 1-inspired - hbj4 - Berlin Strings SINE OOTB except ff- and f-layers off.mp3

Please note that these files are of different loudness and thus cannot be directly compared. That said, if we focus on how dense/massive the sound is, i.e. kind of thinking “how many players are playing here” when comparing the different audio examples, it would seem obvious I think that the example with the fewest remaining dynamic layers sound leaner, smaller, than the first example/the OOTB example.

All this leads me to ask myself this:

Which of these 3 examples is the most correct representation of what 6 (or is it 7?) Second violinists sound like when they are playing Staccatos in Teldex?

(I can’t remember the section size of the Second violins in Berlin Strings but I’m guessing 6 or 7)

Or put differently:

Making realistic-sounding orchestral music with sample libraries is hard!! (esp. with no experience playing in an orchestra or similar close knowledge of how a symphony orchestra sounds)

…

Just for comparison, here’s CSS 1.7.1 playing the same MIDI OOTB:

View attachment Battlefield 1-inspired - hbj4 - CSS 1.7.1 OOTB.mp3

Click to read more...

 

[NoamL]

Well, all the examples have the same number of players: 6 violins.

Berlin Strings is 29 players which is on the smaller side. I have a reasonably up to date comparison table here. https://vi-control.net/community/threads/string-libraries-table-2022-update.85848/

What you are noticing about the mezzo-piano to mezzo-forte example sounding "best" is not wrong, though. You are 100% on the right track!

Orchestral instruments have a large dynamic range but they have a sweet spot where the sound is most characteristic and rich. Sort of like the human voice has a normal speaking volume, where most of the action happens, not shouting or whispering.

John Williams, even in his "epic" moments... when you dive into the score you see he's not asking all the players to perform as loud as possible, but often just mezzo-f to forte. What makes the music pop is the really effective use of timbre, color and clarity of arrangement.

I go over that at the end of this video from years ago (7:40)

 

[jbuhler]

Warning: This is an imperfect test! But hopefully still useful for a certain purpose (more below)

This is Berlin Strings (SINE) playing some Staccatos OOTB (and please excuse my playing):
View attachment Battlefield 1-inspired - hbj4 - Berlin Strings SINE OOTB.mp3

Now, let’s keep everything the same as above, except we turn off the top dynamic layer, ff.
That sounds like this:
View attachment Battlefield 1-inspired - hbj4 - Berlin Strings SINE OOTB except ff-layer off.mp3

Now, let’s keep everything the same as in the OOTB example above (the first example), except we turn off both the top dynamic layer, ff, and the one below that, f.

Remaining dynamic layers are p and mf.

That sounds like this:
View attachment Battlefield 1-inspired - hbj4 - Berlin Strings SINE OOTB except ff- and f-layers off.mp3

Please note that these files are of different loudness and thus cannot be directly compared. That said, if we focus on how dense/massive the sound is, i.e. kind of thinking “how many players are playing here” when comparing the different audio examples, it would seem obvious I think that the example with the fewest remaining dynamic layers sound leaner, smaller, than the first example/the OOTB example.

All this leads me to ask myself this:

Which of these 3 examples is the most correct representation of what 6 (or is it 7?) Second violinists sound like when they are playing Staccatos in Teldex?

(I can’t remember the section size of the Second violins in Berlin Strings but I’m guessing 6 or 7)

Or put differently:

Making realistic-sounding orchestral music with sample libraries is hard!! (esp. with no experience playing in an orchestra or similar close knowledge of how a symphony orchestra sounds)

Go to some orchestra concerts (they aren’t generally that expensive). Note the difference in detail you hear between when the strings play loud versus when they play soft.

I’m generally not an advocate of comparing samples to live playing but will generally point to recordings instead. But there are many things that observing real players in a real room can help improve understanding. Bowing is another facet of string playing that it helps to see and hear in person (videos can be useful but the full effect is somewhat dulled by the sound being channeled through the microphones).

 

[NoamL]

Another example of this is one of JW's most epic moments, the Jurassic Park island theme. Looking at the dynamics, the trumpets are 'only' playing single-f forte for most of it. The real magic is the arrangement, especially how he lines up the trumpets against that wonderful high open Bb again and again, and then in the last iteration of the theme, they burst past that Bb to the high C.

If you know what is idiomatic for each instrument, you can create parts that are soaring but actually comfortable for the players, and that is always the combination that will get the most powerful and beautiful result.

 

[Henrik B. Jensen]

Thanks NoamL and jbuhler.

The opening post is rather messy and the meaning is unclear

What I discovered and wanted to share, I think, is this:

Don’t turn off dynamic layer(s) in Berlin Strings (SINE) because it changes the “massiveness” of the sound / the color.

Just use the library as it comes dynamic layer-wise.

Same for all other OT libraries where you can turn dynamic layers on and off, I suppose.

 

[Henrik B. Jensen]

Oh, I can see you added above that I was actually on the right track in the opening post, NoamL!

Edit:
But to get the right or correct, or realistic, sound, one cannot fade between dynamic layers because doing so changes how “massive” it sounds.

Meaning, if I solo the mf layer above, that’s 100% accurate; that’s what Orchestral Tools got and recorded when they asked the players to play mf.

But as soon as more than 1 dynamic layer is active at the same time, which happens when we use the modwheel for crossfading between dynamics, the sound becomes unrealistically “massive” / “fat”.

The section will sound “bigger” than it actually is, a bit like there are suddenly more players playing than when using only 1 dynamic layer.

 

[jbuhler]

But as soon as more than 1 dynamic layer is active at the same time, which happens when we use the modwheel for crossfading between dynamics, the sound becomes unrealistically “massive” / “fat”

Sort of. You will be faking one thing (blending of two recordings, which won't necessarily fatten the sound, but will often rather blur it) instead of another (real strings move with continuity between dynamics). So if you keep the dynamic layers pure without mixing they will also sound fake because real strings don't jump dynamic layers, nor does the timbre remain the same as volume increases.

So, yes, samples are a compromise formation, and not really real. At best they approximate the recordings that they also are. Yet, you'd think more dynamic layers would help, but that's true only up to a point, because the more layers you have the more layer crossings you have. Blending of layers can be a problem but in practice usually is not, especially for strings since the timbral shifts with dynamics are more subtle than with other instruments. But with more dynamic layers you have more blending or more bumps due to all the layer crossings, and all of that takes careful scripting by the developers and with only 128 points on the modwheel to crossfade among dynamic layers, the regions where each dynamic layer can unfold are increasingly small.