Re your Steinway project, I think that this is the wave of the future.
Re your Steinway project, I think that this is the wave of the future.
I'm not particularly privy to any great secrets, but I
believe that's the direction much of the technology is
headed, including Garritan.
It makes sense. We've got the experience -- the know-
how is there. And now, we also have the kind of (computer)
horsepower that can really handle it, right on the desktop.
When you move to physical modeling, it's an "all things
are possible" scenario, in that you're no longer tied to an
actual waveform... but, rather, a theoretical analog of the
sound that can be manipulated literallly any way that you
We live in exciting times, in this regard. The technology's
moving ahead by leaps and bounds to greater and greater
fidelity to the real sound and behavior of the instruments.
Given Garritan's practically revolutionary Strad (and now
the cello coming along), I'd expect nothing less than a
stunning performance with Garritan's forthcoming
Steinway, as well.
Sampling or physical modeling? I think we'll see both used together in the same product.
This kind of reminds me of the time in the mid to early 80s that analog and FM digital synthesis was paired with the ability to use sampled acoustic wave tables as oscillators--the Roland D-50 as the first best example. Suddenly anything seems possible, sound-wise, and it's musically expressive and useful.
And don't forget that physical modeling of plucked strings is just about as old as digital FM; it's just taken longer to evolve.
I'm sure we'll see an explosion of innovative new methods of synthesizing sound. And the best part is that almost all of it will run on our personal computers.
But don't bet the farm on physical modeling making sampling obsolete. There will always be a place for multisampled waveforms. Here's hoping that Garritan stays on the cutting edge, whatever develops.
Atlanta, Georgia, USA
Music Copyist in Sibelius
Apple MacBook Pro, Mac OS X 10.8 Mountain Lion
Apple Certified Support Professional. I also work with Windows.
So do I, Wheat, and I should have been clearer than "much of theOriginally Posted by Wheat Williams
technology". I've read several wags asserting sampling is dead;
but I don't think that's the case. I'm hardly a credentialed industry
observer, but it seems to me older sound production technologies
generally persist, and co-exist in partnerships and blends with newer
Piano is an attractive candidate for modelling because of its complexity, particularly when developers want to provide an instrument with a broad dynamic range.
I'm hoping that Gary G.'s Steiny will have a true una corda [not sostenuto, as I originally typed;-)] capability.
In the meantime, I like very much that it'll be possible to vary hammer hardness when pianoteq's instrument is unveiled. Lovely possiblilities for tone shapng at the lower dynamic levels.
We shall see....
I'm impressed by what Pianoteq has accomplished so far, but I hope they really keep pushing this for the realism. The samples on their site show me there is still a ways to go there. There are moments of extreme believability but they are largely surrounded by many moments that betray the artificial sound source.
In particular, I find the middle register a little lacking, and anything approaching pianissimo to come out sounding a bit electronic. Both of these come together in a very obvious way at the opening of their Debussy (Fille).
Don't get me wrong, I am excited by the technology. But I think I'll wait to see what v2 might bring to the table.
Well, I have downloaded and tested the Pianoteq demo. At this stage, I still very much prefer the GPO Steinway. For me, the Pianoteq has too much reverb, even when turned off, and sounds like a Yamaha, not my favorite, or perhaps, a Schimmel, also not my favorite. Testing it was a bit klutzy, as the demo has 8 disabled notes. I had planned to upload demos of the Presto of the Moonlight Sonata, one GPO, one Pianoteq. But because of the disabled notes, I decided it would be an unfair comparison.Originally Posted by David Ferris
Originally posted by howardv on PianoWorld,
"I wonder if modelling might work better if they teamed it up with an automated profiler; a tool to analyze particular piano samples and suggest parameter settings to approximate its sound. And perhaps the ability to use different parameter sets on different parts of the keyboard."
Gary asked me to respond here (as I've been doing piano work).
The pianoteq modelled piano is pretty amazing. The piano is really a complicated instrument and to render notes (in real time) with good results, when it's based completely on models is an achievement. There's so many different factors contributing to the sound of a piano note... keybed noises, hammer shank resonances, string resonances, duplex resonances, body resonances, room impulses; even the low notes in isolation with a single string have three separate modes of vibration (transverse, horizontal, and longitudinal) which for each harmonic decay at different rates starting with different gains and (if I remember correctly) are not even exactly the same frequency due to differing impedance at the bridge to different directions of vibration(whew!!). Then consider... to make it even worse, most notes actually have two or three strings (along with the duplex strings that sit behind the main expanse of the string), and so acoustic coupling between these strings comes into play too, to further tweak the pitches and decay rates. The complexity of it is staggering and the physics equations are not for the faint of heart! I'd imagine pianoteq probably didn't model everything possible, but only the most significant contributors to the sound. Still very cool.
Nevertheless, I think you'll be more likely to see Garritan doing physical modelling of brass, woodwinds, and maybe strings, and not piano. For most of the aspects of piano, playback by samples is so spot-on that physical modelling isn't necessary, at least in my opinion. Once a key is pressed on a real piano, the only major change a player can introduce to that note is to stop playing it - there are no crescendos, no vibrato, slides or anything like that which samples traditionally are really weak at doing, and which I think physical models can be very strong at doing. There are some aspects of the piano that do need to be modelled and you'll see that in the Steinway, but it's definitely sample based primarily and modelled secondarily. So I suppose we're doing piano modelling to an extent, but it's not a physical model based instrument. I think you'll see the more heavily modelled instruments from Garritan as brass and woodwind though, and not piano or any percussion instruments either. The neat thing to me about piano work like pianoteq's is the ability to create realistic or interestingly unrealistic sounding pianos while having the power of a synthesizer to change the sound. But when it comes to trying to achieve the sound of a particular great piano in a great hall as tuned and voiced by a concert technician (of which the piano might be the easiest thing to afford day to day in life!), samples seem to me to be the best way to faithfully recreate such optimal conditions.
Thanks Jeff for the explanation. Jeff is our lead person working on the Steinway Piano project and has explored these concepts in detail.
As Jeff pointed out, the sound of the piano is rich with complex resonances and very difficult to model accurately. Samples do an excellent job of faithfully capturing the sonic qualities and subtleties of the piano.
Percussive instruments, or instruments of little sustaining power (which decay after sounding) are where samples shine. With sustaining instruments, tonal aspects are fluid over time and there are relational transitions between notes. With these instruments physical modeling has the advantage. The Stradivari violin is a sustaining instrument and uses a great deal of modeling and fewer samples to achieve its objectives. Bear in mind we are still in their earliest days of modeling.
For the Steinway we are also taking a hybrid approach. The piano is a percussive instrument and we will rely more upon the sample end. We will also use modeling for sympathetic resonances, pedaling, acoustics and other aspects.
Our commitment with Steinway and Sons is to faithfully capture the sound of a Model D and we believe samples combined with some elements of modeling can best achieve this.