Do More Humans mean more Humanization?

[jasensmith]

I think the opposite is true.

I only use the humanize function in solo instruments or small ensembles not more than 3 players.

Why?

The way I see it, the more players playing the more likely their individual subtle tunning and timeing mistakes will cancel each other out rendering the ensemble performance at or very near perfect pitch and timeing.  I know that's probably a bold statement.

Take Orchestral Strings 14 violins.  That's 14 players playing in unison and remember we are usually trying to emulate a professional orchestra with skilled and seasoned performers.

So if 14 violin players play a note in unison, maybe 3 played a few milliseconds early maybe 4 played milliseconds late but the rest played on time.  And most of us layer our strings to simulate the early starts and late stops.  The same idea would go for tunning.

Anyway, I'm just curious to know what other users think of this mentality.

Also, I'm curious to know how others use the humanize funciton.   

[William]

This is very interesting to me and very important for string playing.  The tuning discrepancies of even excellent string ensembles would be much more noticeable if each player was heard separately.   But the overall sound  forms a tone that is beautifully in tune (in a good orchestra).  And the players hear that and feel where they are in creating that overall sound.  Interestingly, the same thing happens in a good concert band, with the clarinets doing a similar complex intonation which is infinitely richer than a chamber or solo sound.  The resonance of a great symphonic band is wonderful and it comes from that complexity. 

One question that arises is if you use all the Dimension Strings how much humanization to apply.  I've been experimenting with how much sounds right, and it is far more than for example a small woodwind group which start to sound like beginning school players at a certain point.  This doesn't happen with the DS and it shows just how far an "in-tune-but-out-of-tune" principle applies with strings (or concert band). 

But it is probably true that with a large ensemble sampled like Appassionata Strings, one will apply less humanization because in a sense it is already "baked in."     

[MMKA]

@jasensmith said:

I think the opposite is true.

I only use the humanize function in solo instruments or small ensembles not more than 3 players.

Why?

The way I see it, the more players playing the more likely their individual subtle tunning and timeing mistakes will cancel each other out rendering the ensemble performance at or very near perfect pitch and timeing.  I know that's probably a bold statement.

Take Orchestral Strings 14 violins.  That's 14 players playing in unison and remember we are usually trying to emulate a professional orchestra with skilled and seasoned performers.

So if 14 violin players play a note in unison, maybe 3 played a few milliseconds early maybe 4 played milliseconds late but the rest played on time.  And most of us layer our strings to simulate the early starts and late stops.  The same idea would go for tunning.

Anyway, I'm just curious to know what other users think of this mentality.

Also, I'm curious to know how others use the humanize funciton.   

I think the same.
I think that the tuning humanization for large groups gives a unnatural result, because players never are playing a little untuned in the exactly the same moment in exactly the same way. The same for the delay humanization.

What I'm testing at this moment is the combination of
- Synchron Strings without humanization,
- layered with SYNCHRON-ized Chamberstrings with humanization (some tuning and some delay), 
- and layered with a solostrings player (I created a ViPRO setting that matches more or less the Synchron
  PLayer). Of course I put some humanization in the solostring instrument also. 

This create as far as I can judge at this moment a good natural lively sound. 

[JimmyHellfire]

It's true that the "humanization" is already there in some large section recordins - Appassionata is a good example. That's why the short notes are not as snappy, gritty and on point as you would get in smaller section string libraries. The long notes feature an audible chorus-y "spread" effect.

I still like to use a bit of humanization even with those samples, especially the tuning portion of it - simply to have a bit of a subtle difference between long notes of the same pitch within a phrase. Even with multisampling and round robin, this is one situation where samples can sound tiring to the ear and overly artificial.

You obviously don't want to have very different sounding samples on one note and in the same dynamic. The sampling itself needs to be consistent, otherwise your sequencing of phrases ends up a frustratingly unpredictable experience. There are some Kontakt-based libraries of other companies where I'm forced to work around (for example exclude certain notes from round robin) the "bum notes" that are in the sample pool.

VSL is always very consistent and even, so that you can trust that a particular note will always sound the same, no matter how often you play it in succession. But the downside is that if your phrase includes a certain pitch multiple times with a similar duration, it can sound a bit strenuous and "fake" because the same pitch keeps on playing back like an alarm sound. Some slight random humanization in addition to altering the velocity x-fade and expression curves can help offset this a bit.

[William]

Totally agree with Jimmy and also have to add that even though it may be improper I am still using a lot of humanization even on the larger ensemble sounds.  Theoretically it seems wrong but to the ear it is better.  But I am a pariah with the amount of humanization I am using.  Many people find it sickening.  

[jasensmith]

@Another User said:

VSL is always very consistent and even, so that you can trust that a particular note will always sound the same, no matter how often you play it in succession. But the downside is that if your phrase includes a certain pitch multiple times with a similar duration, it can sound a bit strenuous and "fake" because the same pitch keeps on playing back like an alarm sound.

Come to think of it you're spot on about that.  I've been using the stop offset to pump some variety in those same pitch passages but I hadn't really thought about using humanization and I think you may have changed my mind about using it in large ensembles.  So do you just keep the tunning at about 15 to 20 cents or keep the tunning fader around 15 or 20?

The key to Humanization is the same for Vel X Fade and Time Offset or Legato Blur it's just the right balance of the effect.  I think if it starts drawing attention to itself then it's too much.  However, I have been known to go a little heavy handed with solo performances and still have been surprisingly satisfied with the results. 

Thanks Jimmie, William and MMKA.  To be honest, I didn't really expect too much out of this thread but in the span of just four posts I've already changed my point of veiw on Humanization..

If anybody else out there disagrees or has some other contribution to make on this subject I'd love to hear from you.  

[JimmyHellfire]

@jasensmith said:

So do you just keep the tunning at about 15 to 20 cents or keep the tunning fader around 15 or 20?

Generally I have the global humanize scaler all the way up, so that it's always doing something. I then adjust the tuning fader depending on the piece I'm working on. I like to keep it around 35 and dial back if it's too much. Sometimes I'll have it below that, but if there's a note where I really want to hear a change, I map the tuning fader to a CC and ramp the value up in the corresponding MIDI lane in Cubase just for that note or passage.

[jasensmith]

Thanks for that tip Jimmie.  I'll try that next time along with some experimenting.

One other question if you don't mind.

Do you employ this technique for all ensembles?

What about the strings?  I assume you layer your strings with the smaller ensembles or solos, do all layers get the same treatment or do you something different for each layer?

Okay, I guess that was actually three more questions but thank you for letting me pick your brain on this.

[JimmyHellfire]

I do layer strings most of the time. If I'm for example layering Appassionata with Solo Strings, I would finish the Appassionata performance, then copy the MIDI events to the Solo and listen back and adjust things like velocity crossfade etc. It's two entirely different sample sets that behave differently, so one shouldn't just copy stuff 1:1 and leave it as is.

I however generally do not re-do humanization, unless something sounds wrong. Even if you use the exact same humanization settings for both libraries you're layering, there's gonna be enough variation either way when using the "in tune random" preset. So I think just leaving it as is and letting it do its thing is OK most of the time.

[jasensmith]

That's what I figured.

Thanks again Jimmie.  This has certainly been an eye opening experience for me.

[Paul McGraw]

I did a lot of work recently on the Dvorak 9th. I was working to match the sound of a particular LSO performance on CD. Based on my experience, I agree with most of the comments above. I do find that even in a top group like the LSO, fast passage will have more timing and intonation variation than slow passages. To simulate that effect, I use CC controls within the work to vary the amount of humanization depending on the tempo and complexity of the passage. I was pleased with the result.