Been wondering about and fighting with this for a while.
It's kinda hard to explain though.

Old school synths play with a velocity sesitiveness that seems natural and normal and smooth. How hard you hit your controller defines how loud the sound. Modern day vsti's don't accurately. Some are closer than others. 


The problem here I think, is that the vsti's use a recorded 'soft' sample for lower velocities and a 'louder/harder' sample for higher velocities. Makes sense I guess till you try to play it with any kind of nuance. Then, it doesn't make sense. As it would seem that the sample programming of soft to loud is fighting the keyboard's/finger's soft to loud.

A vsti with only one sample layer seems to respond more accurately to velocity from my controllers than one with multiple sample voicings. Multiple sample voicings may "sound" better due to more variations. But, I think the whole mapping concept is skewed.

It would seem to make more sense to have the sample pool of a particular vsti sound normalized. No? 
That way the playing field is kind of leveled back to the keyboard/player.

Or, maybe I'm missing the point of multilayered vsti's in that they are not meant to be played but are meant to be programmed.

What do you think?

A big part is from your actual controller.
Some have a fairly high velocity compared with others.
I use my Yamaha XS8 (88 weighted keys ) which I find quite good.
My Akai and Keyrig doesn't feel as good...

Well, not all old-school hardware synths have the same velocity response, and many of them allow(ed) the user to select one of several different velocity response curves (e.g., linear, concave, convex).  Most keyboard MIDI controllers these days have assignable response curves as well, as do many soft synths.  By default, these are usually set to linear, which is probably not the best match for your particular playing style.  Depending on whether you're accustomed to playing weighted (e.g., piano) keys or semi-weighted (e.g., synth) keys, you'll probably find one of the concave or convex response curves a better match and more expressive for your playing style.  Start with your keyboard, since that is the common demoninator.  Then, change the response curve in your synths (if allowable and as necessary).  The DCAM Synth Squad synths, for example, allow you to save a custom velocity response curve for each patch.

Also, keep in mind that the velocity controls a lot more than just amplitude.  It's often used as a modulator for other parameters (e.g., filter cutoff, LFO rate), so the differences that you are hearing, or differences in expression that you are experiencing while playing may be attributable to much more than just volume.

Depends on whether you're talking about synthesizers or sample players or hybrids.

With a synthesizer, velocity response is programmable. In a full-featured synth, velocity is just one more modulation source that can be routed to many things, so that you, the programmer decide exactly what different velocities will do to the sound. Including how they'll transition from soft to hard, which is referred to as the "velocity curve".

Sample players don't have the same degree of flexibility, but make up for it by being able to use different samples for different velocity ranges. For sampled pianos, you might see as few as 3 or as many as 20 velocity layers. Badly-designed or too-few layers make for an instrument that doesn't feel right.

In either case, an important feature of any VSTi is the ability to alter the velocity curve from its normal linear shape. Sometimes, keyboard controllers themselves offer this feature, which is nice to have because then you can customize the response no matter what instrument you're driving. Otherwise, it's up to the instrument vendor to provide that feature - and not all do.

Thanks guys.

I have a Roland A-30 keyboard controller (ancient) and a Korg padkontrol.
I get just as frustrated on both.

The A-30 does have 3 velocity curves. I always thought they were 'light medium high' or something. But yes, they could be different shapes. The A-30 acts exactly how I would expect/want it to act with something like my JV-1080 synth but not with stuff like Kontact.

Bit, Yea, I guess I'm talking more about sample playback engine vsti's. 

Still though, doesn't it seem a little odd that say a Kontact drum or piano patch with say 20 layers per hit is programmed the way it is? Being that, the layers ARE at the volumes at which they are recorded, NO?
So, it's like a double negative or something when you play them with a velocity sens keyboard??

I mean you play lightly which triggers a sample that is lighter in volume already.
Ahh. SO, With no velocity curve on the keyboard maybe the samples would respond correctly??

Hope that makes sense.

Getting the levels right for multiple sample layers is tricky. Some library developers go to a lot of effort to get it right, others don't seem to even understand the concept. 

But no matter how hard you try to make those transitions transparent, there will always be an audible change from one sample to the next as you cross the threshold of one velocity layer to the next. 

That's touted as one of the benefits of modeled instruments: continuous velocity response. I have heard at least one instrument that really does accomplish this, one of the newer Roland pianos. It has no velocity layers that I could detect.