Piano World Home Page Forums Our Most Popular Forums Digital Pianos - Electronic Pianos - Synths & Keyboards New Pianoteq Piano

Nice play, Yuri. And thanks for the demo.

Yeah, nice playing Yuri but I'm afraid that for me, this suffers from the same malaise of all the Pianoteq pianos, which is that I just don't hear enough "wood and wire" in the modelling.

Hmm, if you play it that much you should definitely update if you have not already.

But depending on your DP there is the possibility that the partial pedal on Vintage D (or any software, including PianoTeq) may not work the way you want it. Progressive DP pedals only transmit a few velocities (that way they don't completely swamp the DP with signals). For example, you might get velocity 0, 15, 30, etc. The software defines a range of pedal midi levels that it considers to be the half-pedal rannge. Now if the software is considering MIDI levels between, say, 50 and 60 to be the half pedal range, it's possible that your pedals don't actual transmit any levels in that range (or it's hard to get there).

For this reason, software pianos should have tunable parameters for the level and width of the half pedal range. Most do not, although I think you can translate the MIDI velocities using a third party tool or maybe even Kontakt, just as we might remap the velocity curve for keys.

Anyway, it's completely possible that you are having problems with half pedal while others do not because your hardware isn't sending the velocities Vintage D expects. Just FYI.

Personally I have always found Vintage D partial pedal and repedalling to work swimmingly.

I have often noticed that while you play, you do not notice the many nuances, but then - in the record. Pay attention to a whole other thing - the adequacy of the reaction, etc.

Well your statement referred to ALL sampled pianos. So your statement was pretty sweeping. You can state any opinion you want, but expect some criticism when you make a statement as general as this one and then say you have experience with only ONE sampled piano:


On top of that your specific criticism of the Vintage D is bogus. It's pedaling works great unless you have never updated the original release code. So you took a problem that doesn't exist anymore (or is specific to your hardware and correctable as discussed by gvfarns) and generalized it into a problem with ALL sampled pianos.

If you prefer Pianoteq over the Vintage D, or any other sampled piano you have tried, that's great. You can state any opinion you want, and there will never be agreement on everything related to digital pianos. But we try to help people here and a lot of folks new to digital pianos visit looking for information. So if you make sweeping generalizations that don't apply to all products expect to be challenged, and if you make criticisms of specific products that aren't valid issues for other users than expect that to be corrected as well.

In the spirit of trying to help, make sure your Vintage D has been updated to the latest version. The updater process with Kontakt and Vintage D is a bit confusing and has fooled a number of users into thinking they have updated when they really haven't (there are other threads on that subject). Make sure you are actually playing the latest version. If you still have problems pedaling with a pedal that supports multiple MIDI values, then read gvfarns comments. I don't know if that MIDI hardware issue is very common, but it is correctable with a software utility.

I like pianoteq for the natural response to my playing and resonances and natural decay. I don't like pianoteq for its attack phase. Something artificial and too " woolly / thick " to my ears. From The early pianoteq versions I always missed the 'air' it he sound getting to thick and tiresome after a while. Hard to pin down on exactly what causes this. Overall I think pianoteq is not bad at all, escpacially if you tweak it to your liking. And it's only a few hundred bucks. Can't by a real grand for that...

I don't know guys... This new piano sounds very nice to my ears.

Perhaps it's that I'm too used/tired with samples that I'm looking for something different. Perhaps it might be that I'm kinda like a 'fan' of pianoteq, and got the models as a prize for a competition long ago... Nicklas has been kind enough to keep supplying me with the new models and I keep liking what I hear.

I will admit that if I compare the older models to this one, there's a lot of improvement! And in regards to playability there's nothing like pianoteq. And I do have (had) the old Giga piano, Ivory I, the Garritan Steinway and a few more... :-/

Those are all first generation sampled pianos. They don't play well like the current generation does. Well, actually I've never tried Garritan, so that one might actually be good. I find Ivory 1 basically unusable, for example.

People make a bigger deal about sampling vs modeling than makes sense. The basic tone of the piano can be sampled or synthesized. That much is true. Everything else must be modeled, whether it be in PianoTeq or in a "sampled" piano. A sampled piano records 13 or 20 levels or whatever to get the basic tone and interpolates to get the other 117 or so available. PianoTeq synthesizes all 127 of them. Beyond that there is no difference.

Now, you may think PianoTeq has done a better, more complete job at modeling all the nuances and you have a valid point, but it makes no sense to draw a line between pianos with and without modeling because all current and future pianos have modeling in them.

The only remaining question is whether it makes sense to synthesize the piano tone or construct it by recording/interpolating. Based on what I see, we will continue to see the latter sounding better for a long time. But the modeling aspects of all these pianos will continue to improve until there can be no more gains in playability. Actually, we are nearly there now.

Sampled pianos have modeling in them? Huh?

Interpolating between, say, 20 sample layers is not modeling at all. And I've seen nothing to suggest that they even do such interpolation. If a piano library has 20 sample layers, then it has 20 sample layers, not 127.

It may have 127 volume levels. But shifting among them is simple audio processing, not modeling.

Sampled piano libraries have very little in common with modeled ones.

Well, define modeling however you want, but semantic changes don't affect the underlying similarity.

There are 127 distinct timbres if the timbre is interpolated between layers (which admittedly, is only true for some pianos, Ivory being one...this is one reason that people who say they can hear the layers in Ivory are so full of crap). Sympathetic resonance sounds, etc, are computed in every piano using an algorithm I'd happily call modeling. If it's a bunch of code and not a recording that produces the sound, it's modeling in my book. Further, some of PT's sounds (like hammer noise) are implement more like a recording than modeling because they are pre-computed and included as audio files.

For marketing purposes, the PT people talk about their modeling as if they created a piano image using CAD or something and then all the details just fall out. No, they had to code up every behavior that's there, just as a sampled piano does. Strip away their code for the various resonances and reverbs and you would get a tone that sounds like a simple string being stuck, just more artificial.

Now if the sampled piano neglects to model some particular aspect of the sound (i.e., they don't code up some algorithm to produce the sound) that's a shortcoming of that particular piano's implementation, but it doesn't mean the things they did code up and include are done in some way that is completely different from that the PT people had to do.

Edit: reference for ivory:synthogy's ivory page. Notice the blurb about sample interpolation technology and also the bit about harmonic resonance modeling. Actually, you know there is what I call modeling in your sampled piano because of all the many things you can change about the timbre and behavior. It's even more clear in hardware digital pianos.

What modeled pianos can really nail is sympathetic resonance. Stimulate those string models through the soundboard impedance function and you can really get some resonance going.

I don't think producing 127 different timbres at 127 different volume levels is really the issue. In fact I believe that is a fairly insignificant part of emulating a realistic piano. There is at least one sampled virtual piano that claims to have nearly done that however. Vienna Symphonic Library's Vienna Imperial claims to have up to 1,200 recorded samples per key (damper pedal up samples, damper pedal down samples, release samples, sympathetic resonance samples, repetition samples, multiple mike positions, etc. at up to 100 velocities per key) with about 70,000 samples total (about 500 GB of uncompressed samples) using the Bösendorfer CEUS technology for that accomplishment. However, despite being (one of?) the most throughly sampled virtual pianos it lacks simple half pedaling and repedaling behavior, which is accomplished through (non-physical) modeling on other sampled pianos. As a consequence (I believe) it is rarely (if ever?) mentioned here with other favorite sampled pianos that include those features, such as the Galaxy Vintage D (13 layer sampling) or Ivory II pianos (16-20 layer sampling). (I personally haven't purchased it because it lacks those pedaling features.)

We probably need to distinguish between physical modeling (used by Pianoteq I'm told) and using sample-based DSP modeling to emulate the sound of a various piano characteristics. I'm sure someone that worked in this field could supply a better description, but I would describe it like this. In both cases the intention is to emulate an acoustical piano, i.e. produce the sound and playing characteristics of an acoustical piano. But in physical modeling the goal is to describe the mechanical and acoustical sound producing mechanisms from their dynamic and materials properties as mathematical equations that can implemented using DSP techniques. It's an overall transform from MIDI velocity to piano sound output. However, whether Pianoteq maintains a strictly physical model for everything, or also applies additional non-physical modeling for some functions, is known only to them.

In sample-based modeling a core set of MIDI velocity to sound samples are recorded and then a model is created to transform those primitive samples to piano sound output characteristics (which should include release behavior, simple sympathetic (string) resonance, sustain pedal resonance, half pedaling, repedaling, etc.) The designers of sample-based pianos have chosen to produce those characteristics using either sampling or modeling techniques that include additional samples (release and/or sustain resonance samples) in some cases, or through purely DSP modeling techniques in other cases. For instance, Ivory II and Vintage D both use release samples, while Vintage D uses sustain resonance samples while Ivory II uses DSP modeling for sustain resonance. Both use DSP modeling for sympathetic (string) resonance, and both use modeling to create half-pedaling and repedaling. Both use other DSP modeling techniques to implement various voicing options, and include convolution reverb processing, etc.

The point is that the samples are just the core primitives on which the overall virtual piano is created. Non-physical modeling is applied to the samples to create the final result and the differences in those techniques are differentiators between sampled products. It seems to me that the most distinguishing characteristic between Pianoteq and the latter two virtual pianos that I mentioned, is that those sampled pianos, with their measly 13-20 timbre layer primitives, actually sound like the Steinway pianos that are attempting to emulate, while in my opinion Pianoteq has never sounded like any actual acoustic piano that I am familiar with. Until the day comes that Pianoteq can do that, it isn't satisfactory to me.

Makes sense, Macy. As both of us pointed out, the method of producing the fundamental tone is completely different between PT and sampled pianos. I guess the succinct way of saying what I said previously is that I believe the rest of it, which you describe as the more significant part of emulating a real piano, is very similar across software pianos, and it can be described as modeling in all cases.

Though you do rightly point out that some software pianos go hard core into sampling lots of different cases (VSL is an apparent example) rather than using a modeling approach. That doesn't seem to be the approach of the most popular pianos in this forum, though.

I have also tried the QL pianos (gold not platinum, but what the heck, it can't be all that different) and I gave a go to the VSL Bosendorfer in Frankfurt at some point!

I still backup what I said... :-/

I think there is. I'm not sure about all the technical stuff, but in the end the one is jugsawing different recordings, while the other is creating them one by one. Even with round robin and 120 velocity layers it's still not the same.

Again I'm not sure how you mean it, but for the QL pianos and the Garritan one (which I happen to also adore) I think that if there's modelling that's limited to some sympathetic resonance...

Other than that, if you compare the 270 GB of QL pianos vs the 21 (?) MB (not GB) of pianoteq it's as valid as it can sound (especially if all the new sample libraries (from EW and VSL, or perhaps other companies as well), require you go potentially buy a SSD drive to run them from...

Perhaps... perhaps...

The one thing that I like about pianoteq is the fact that they keep modeling old pianos... This is fun! ^_^ Other than that all pianos and sample libraries and modelled stuff have found their way into my productions and I'm fine with everything I do (and so are my clients)...

The QL Pianos are similar to the VSL I discussed earlier, in that it also attempts to emulate acoustic piano resonances by separately sampling notes with the sustain pedal up and the sustain pedal down. It also samples repetitions (pedal up and pedal down) separately from non-repetitive notes, and even samples staccato notes separately. In other words, it tries to use brute force sampling (a flawed approach in my opinion) rather than modeling to emulate various acoustic piano characteristics. Also like the VSL it has no half pedaling or re-pedaling. It's just my opinion, but I consider the QL Pianos old, inferior technology and rank them at the very bottom of my virtual pianos (by a huge margin).

I don't understand your other comment. If you are saying Pianoteq is as good as it can be for a 21 MB program I doubt that. I would expect it can still get much better as it improves its models. Of course the models may grow in size someday, but who cares? It could grow by a factor a 10x or more and no one would care. The real issue is whether it is currently constrained by CPU/DSP processing power or by insufficient research to improve the modeling? Either way, it should become better in time.

On the other hand, if you are saying that sampled pianos are less desirable because of their size or their storage requirements, I'm not concerned about the cost of an SSD or bigger hard drive at all compared to the price and size of an acoustic piano. It's just a non-issue.

For me, the real value of a physically modeled piano will come when I can simply dial in a set of parameters to create a convincingly realistic digital emulation of any piano I want after someone else performs a few physical measurements on that acoustic piano. At that point we can have a realistic new virtual piano in a matter of days or weeks rather than years to get a new sampled piano. We can have a new Estonia or Schimmel or Shigeru every few weeks instead of another Steinway D or Yamaha C7 every few years. That's what I'm looking forward to in virtual piano technology. The realism, playability, and computer requirements of state-of-the-art sampled pianos are satisfactory right now. The long wait between models and cost of production (which creates a limited variety of models) is the problem that physical modeling could fix in the future.

Another thing that modeled pianos can really nail is playing repeated notes with the pedal down. The hammer imparts more energy to the already vibrating string, but also removes some in a rather random, somewhat frequency dependent way. Sampled pianos can give you something like the latter with round robin, but they can't do the former. In the digital drumming world the effect is known as "machine gunning".

Yeah, you know, this has never bothered me but it has been brought up before. I suspect sampled pianos could include that behavior as well. I don't see what would prevent them. Probably current models do not because they are made by small teams without much money or mathematical expertise--mostly they are studio people. The PianoTeq team may have little money (I don't know) but math savvy is basically what they have.

@Dewster: Sampled pianos already model that, by simply overlapping repeated notes with multiple voices. You can hear the effect even in run of the mill digital pianos. It's nothing fancy at all - the random phase differences between the overlapping voices creates the subtle changing timbre. Sometimes, it may simply be a bit of overlapping of the release portion of the voices overlapping with the attacks of the new voices - whatever the case, this simple method works well.

One software instrument that does NOT do any overlapping (or, it seems, modelling) is the Lounge Lizard electric piano software. You really notice that it's not there - it sounds very dead. Initially, it DID model it, however every now and then repeated notes sounded far too loud. I reported this problem, and I get the impression that they simply disabled the modelling. When they did that, they should have turned on voice overlapping, IMHO.

Greg.

I agree and I don't fancy the monster approach in the sampling world actually. I largely prefer the idea of sample modelling (the trumpet, Mr. Sax, etc) which is a true hybrid of modelling and sampling.

I'm saying what yo mean in the second paragraph, but I disagree that this isn't an issue. It doesn't depend on your resources and when a single set of instruments (Hollywood strings, or the whole lot of the pianos) request that you also get, btw, a new hard drive and a great one to run successfully it causes a few issues... And the comparison with a real piano is non existent at this moment. Because of many factors, not only that the digital instruments can't reach the 'realism' of the 'real' thing. (which in the end, if you ask me, it will end up as a digital recording, thus arrays of numbers as well).

Fair enough and I'm pretty sure that pianoteq is headed that way little by little. Problem is that exactly because all the products we're talking about need a quite big capital to start of (monetary I mean), it also means that they need to be commercial products and by that sentiment, sadly if you ask me, the need for a Schimmel or a Shigeru piano is so little in the commercial world (that which is fine with a Steinway or a Yamaha and a Bosendorfer and get it over with) that there's little chance anyone will care to do that.

Sorry for losing this thread