Piano World Home Page Forums Our Most Popular Forums Digital Pianos - Electronic Pianos - Synths & Keyboards Sample vs model technology

No - they combine modeling, e.g:
Yamaha - Virtual Resonance Modeling
https://sg.yamaha.com/en/products/musical_instruments/pianos/clavinova/clp-675/index.html

Same with Kawai and Casio (and older Rolands).

Curious. Virtual Resonance Modeling sounds like a trademark. I wonder what it really means?

Looks like no one understands that if computers can create dynamic atmosphere of entire gaming environment interacting with a player, including visual and sound parts, it is not a problem for CPU to model just single musical instrument.

The audio in a piano VST is much more complicated than the audio in a computer game.

The visuals make use of the graphics card to do calculations, so that's a different matter entirely.

It's just the pseudonym Yamaha uses for the various resonances: damper resonances, string resonances, body resonances, etc.

Games almost all utilize samples for sound effects and music.

Virtual Resonance Modeling is in fact just modeling of the resonances They are not real (recorded, sampled) resonances, hence they are virtual. Did I say something new? What's not clear about that?

BTW, IMO the best of both worlds is exactly what Kawai and Yamaha do: sampling + resonances. Which is why my N1X is more playable than both the Garritan CFX and Pianoteq. The only problem is digital piano manufacturers still use short samples and looping. If they applied the VRM on top of VSL libraries or even Garritan CFX, it would be a heaven! I even proposed to Philippe of Pianoteq in one of our email communications, that they may collaborate with some of the sample libraries to give them just the resonance modeling. he didn't answer to that unfortunately. I guess they don't want to have anything in common with other competing companies, let alone sample-based and that's understandable. But IMO that would be the ultimate realism we all strive for.

CG, do you know if the expensive Pianoteq allows you to turn off everything but the resonance modelling? It seems like it would be a very popular with way to use it.

Computers certainly can, no big deal for them, really. Mankind has not yet found a way to
a) understand the math sufficiently
b) code the math.

Esp not when there seems to be an obsession about making VST run on potatoes and toasters in order to boost sales.

The samples are stored in either raw or processed form and then mixed and transformed through DSP algorithms. If we call these algorithms "modeling" doesn't matter. The end result on Yamaha's end is a clean, natural and responsive piano sound one would expect from an advanced hardware synthesizer. And those algorithms are made for specialized DSPs, so they don't run on general purpose PC CPUs.

PC sample libraries are behind, because they are mostly just playing back raw samples. This is 1980s technology, even with hundreds of GBs of storage. "Pure modeling" synthesis is just as outdated, even when slightly more advanced than 2-operator FM synthesis from the same era, it still falls short, just like EPs with pickups did with emulating an AP.



This is the price to pay for the hardware synthesis you like so much. You can either play back huge samples - or do lots of DSP work at the price of having smaller samples. You can't have both.

Audio in most (if not all) videogames is sampled. It's very easy to use sampled sounds in videogames, because you don't have the problem of the timbral changes (based on velocity) and all the various resonances (sympathetic resonances, dampers, body, aliquot strings, undamped strings, etc.). A piano VST has to read many many high-quality samples in real-time (even hundreds). And when I talk about real-time I mean that the output must have a latency of max 7-8ms. This means that a big amount of system power will be used for reading and playing all these samples (+ applying some resonance effects, reverb, etc.). A modeled piano engine can be less or more demanding. It depends on the type and complexity of the algorithms.

In videogames, the number, length and quality of the samples used for FX is much smaller and usually videogames don't need a big polyphony. I'm sure that if you used an high-quality VST piano engine in a videogame, the videogame would have a cr@p framerate (or the framerate would be good but the audio would be full of clicks and artifacts... It depends on what is given the highest priority: audio or video).

I think that's not possible. Most resonances in Pianoteq are intimately part of the note generation algorithms so, probably you can lower or turn off some resonances, but you cannot apply them to an external audio source.

That’s a great idea! I own the Pro but haven’t tested it. Will try to find some time and test it.

Video games play back uncompressed digital samples, as CPU power is a scarce resource on game consoles, while storage is at a surplus. These samples are either prerecorded dialog voice-overs (90+ % of the sound data), premixed music soundtrack or pre-processed FX effects.

The remainder is just playing back the right sample at the right time and panning it to the right place in the stereo or multi-channel mix with the help of the hardware codec. There is no actual software synthesis technology found in video games, so the comparison with DAW plugins falls short.

A convolution algorithm that applies, for example, a wooden body resonance effect to a sample doesn't need a more powerful CPU/DSP if the samples of the notes are longer. It needs more processing speed just if the "impulse response" data is longer (i.e. to get a more accurate resonance effect).
For now, I'm not able to think to a resonance effect that requires more CPU/DSP speed if the samples of the notes are longer. You certainly need more storage memory...

DSPs are distinctively different from CPUs and are very good at some things, CPUs struggle with: for example processing and transforming many samples in parallel with very low latency, which is how they reach high polyphony numbers and a rich real time sound as a result (see Kawai's pianist mode using four DSP slots for each note played).

OTOH CPUs are good at processing bulk data including hundreds of GBs of sample recordings, which is what PC software instruments are about. But once it's not about raw playback and actually doing something with the data, latency quickly becomes a problem. That's why they aim at passing it through as quickly as possible, with very rudimentary processing of for example pedal input.

A standard PC CPU is not powerful enough to implement something like "Pianist Mode" (or Yamaha's equivalent) on there and this is why it doesn't exist. Trying to play back 256 samples at the same time from huge 100+ GB source will quickly run you into memory bandwidth (and latency) limits, but this is exactly what digital pianos do with a much smaller sample source at real time latency.

And that is also the only reason, why Pianoteq can do something more than those PC sample libraries: It generates its artificial tones on the fly with a simple algorithm - saving the bandwidth for the samples and paying the price of only sounding similar to a piano, but not quite like one. It's a quirk, not an advantage.

Casio also do it in their GP series.

That's not what was suggested.

Sorry if I've misunderstood, maybe did he mean to overlap resonances from Pianoteq with a piano sound from another VST or from the digital piano engine itself? But I don't think Pianoteq allows you to disable the sound of the notes and play just their resonances...

That's good I read the proposal-just mute the note itself and play everything else. Even if it's not possible with the public build it sounds like it would be intriguing as a developer option.