OK, this is going to be a simplification so bear with me. 88 keys => 88 samples, taken for 20 seconds each, and saved, compressed to a 320kbps mp3, so about 1 mb, 88x -> 88 mb, then 10 volume levels -> 88x10 = 880 mbs. Other sounds such as pedals can be added, they wouldn't add much to the size. So for arguements sake lets round it to 1 GB, so how come some piano samples such as ivory will come with 10 GB samples? What is taking up all the space? Just ridiculous.

And can sample based pianos really be the future? There is still alot of math involved to blend all the sounds together even with samples, it wouldn't be much trouble to start off with a purely artificial sound and model it to get a great piano sound.

I hope physical modeling takes off, then I can get a piano and change the sound however I want, so if I want a bright or mellow sound I just need to turn a knob, no need for the DP company to sample a different piano and store it in a multi GB fileset. And I don't think sample based pianos can ever come close to the complexity of modeled piano sounds played together in a piece of music.

And so history repeats itself on this blog again...

I think quite the opposite. Sample-based pianos achieve the complexity of real piano sounds because they are a recording of real-piano sounds. Modeled sounds are just a mathematical simulation.

As for the multi-gigabyte size of sampled pianos ... Is that really a problem? I remember when a 5 MB (meg, not gig) drive cost $5000, and 1 MB of RAM cost $3000.

Today, a 10 GB piano sample set occupies about 50 cents worth of disk storage. Expect that to be about 5 cents some time real soon.

I believe that timbre changes depending on how hard the hammer hits the strings. So it's not just about amplitude, it's about tone, and how that changes with velocity. So essentially there needs to be many samples for each key, which in turn creates large samples.

Sound is extremely complex, and the human voice being the most complex soundwave. Acoustic guitar is probably more complex than the piano.

I don't think it would be wise to use a lossy compression such as mp3 for sampled pianos. Technically speaking, I believe there is audio lost when converting to mp3, even at the highest setting, and some people claim they can detect that with their own ears. Though I can't tell the difference between 320kbps mp3 and wav, I would guess that if you recorded with a DP or VST with mp3 samples that the more processing that would be done with it, the lower the audible quality of the recording would become.

For example, say if you recorded a performance of an mp3 piano and saved it as wav, then converted that wav to MP3 to share your recording with friends. I think maybe in that scenario you might notice the audio quality difference; and I think there would be a loss in the actual audio data from the original recording of the piano samples. So you would be gradually losing a part of the original piano sound each time you convert.

So that's the reason I think the libraries might have to be so large.

But, I do stand with you in your hope that modelling progresses even further. I love the sound and expressiveness of Pianoteq 4, and hope that the next version sounds even better.

Adak,
1, Piano libraries do not use MP3 compression, it would slow thinks down considerably and affect the quality of the sound - there is a lot of filtering going on between loading a sample up and playing it, so you want the sample to be as artefact free as possible.

2, You forgot release samples, una corda samples, resonance samples, different pedal down samples, re-pedal samples , hammer and string noises, etc. In the Galaxy Vintage D specs they state: Over 2000 samples in 24Bit (10GB/5GB with sample compression), which makes 22 samples per key.

3, For Ivory pianos (and some others), multiply the sample number by the perspective count.

4, Ad future: Let's wait and see. Right now there are many sample-based VSTs with beautiful sound, and just one modeled library, so modeling definitely isn't easy (I read on this forum that there were other, unsuccessful attempts)

5, Ad modifications: You can change the character of Vintage D sound quite considerably, just by turning the knobs.
On the other hand, in Pianoteq, while in theory there are more ways to change the sound, you cannot achieve everything either, otherwise there wouldn't exist paid plugins, like the new Bluthner.

Me thinks the future will be a combination
of samples and modeling in one box.

The samples provide the actual sound and the
modeling comes in to emulate the resonance and
other complex nuances.

I am so glad I have unsophisticated tastes! I just play/listen to the music, not the instrument . . .

Roland takes that hybrid approach. Although the V-Piano is (I believe) all modeling, the other SuperNATURAL pianos (RD-700NX, FP-7F, etc.) use a combination of samples and modeling.

If used properly, I think its a good thing; as long as the large memory is justified

That's already how it is done with Ivory II, which uses modeling for string and sustain resonance. The Vintage D still uses samples for sustain resonance.

Then why waste money on an expensive
DP when this will do it for you?

As others have pointed out, there are many samples taken per note, to capture the different timbres at different velocities. There's a velocity layer war going on kind of like the megapixels war or the megahertz war. More velocity samples can be a good thing, but there are other things that matter, which are often more important than the raw number of samples.

Some software pianos use lossy compression, others use lossless. I'm not sure I have an opinion on whether they should use lossy compression or not.

As for modeled pianos, I think it's fair to say they've taken off. PianoTeq appears to be pretty successful. And Roland's SuperNatural engine appears to use modeling of some sort for major portions of the sound, to say nothing of the V. Sampled Modeled pianos have a real drawback in that they have a hard time sounding really, really real. They perform great, and their sound is improving, but there's a syntheticness there that seems to me to be far from ready to replace samples.

I think you meant "modeled."

Ahem, modelled.

Both are correct. modelled = Canadian/British, modeled = American.

The one 'l' looks wrong to me but I'm Canadian.

Yes, I know, I'm just having pop at VP's British roots.

Also, I believe:

modeled : America
modelled : Rest of the World

Cheers,
James
x

Oops, sorry for the goof. Fixed in the original post.

Yeah on modelled vs modeled, I go back and forth on the usage. By default my firefox was set up for New Zealand English for some reason and it corrected it to modelled enough times that I got used to it that way. But I have now corrected my language settings and I'm trying to retrain myself to write modeled. Same issue with pedalled and pedaled.

I'm way too dependent on automatic spell checking.

It's not your fault, English is a crazy language. When I was young I used to win class spelling bees, now I can barely write a sentence without making a mistake. All the rules and exceptions blur together at some point. Thank god for spell check, or people would think I'm even more of a moron than I really am.

I was brought up on British English, but on this forum (unless I'm replying to British posters), I try to use American spelling (or should that be speling? ).

Unfortunately, Americanisms are starting to impact on my 'normal' life and I'm begining (sic) to spel program instead of programme, behavior instead of behaviour etc, etc.....

Is there any particular reason why?

James
x

As a ten-year-old I was furious when my teacher corrected my "travelling", saying it should be "traveling". Hmmmmphhhh! I know the rules, I thought. There should be two ells! (I was wrong.)

Becauz I try to be all things to all (wo)men everywhere .

And because this is an American forum, and when you're in Rome......

You were right, according to our Queen Elizabeth II .

Well, Frank (the owner of PianoWorld) is America, and the server that hosts PianoWorld is located in the US, however I don't 'feel' that this is an 'American forum' any more that I feel the Nord forum is German.

James
x

Or Swedish, even...

The Nord forum is owned and run by a German chap, the server is based in Germany.

James
x

You should stay away from software that uses lossy files/compresion if you want to achieve professional recording.

Just for clarity - Vintage D use lossless compression, so everything's fine there. I also think that most softwares worth considering will by default use lossless, because, really, there's no justification doing otherwise.

Hey! The Keys between that thing are evenly spaced!!!!!