Topic by EvilDragon

  • Avatar
  • EvilDragon
  • Member
  • Offline
  • From: Croatia
  • Registered: 27-06-2008
  • Posts: 3,994

Topic: OT: Reflection and discussion topic - in 5 years: modeling >> samples?

Yes, this is the discussion I started on several forums.

Guys, do you think in 5 years, going with current tempo of advancement in physmod and CPU power/multicore technology, we will be able to have modeled instruments that are several times more complex, and with that genuinely insanely close to "the real thing"?

If you watch the advancement Pianoteq took, in 3 years from v2 to v3, there have been some huge leaps in functionality and sound. v3.5 Pro just enabled tapping into that functionality with even more power.

Take a look/listen at WIVI. It sounds wonderful. It's constantly updated. It can model about a 100 SINGLE instruments in ONE room, completely separately positioned and directioned, on a 2 GHz dualcore CPU! Imagine what an i7 can swallow! WIVI is great as it is, but it also has space to improve - both in articulation number and timbral precision. Not to mention Arne is probably working on modeling stringed instruments, so WIVI is probably gonna be a modeled orchestra in few years.

Next, take a look at Synful Orchestra. Amazingly different synthesis there. Amazingly expressive. A few niggles to sort out (splicing precision, getting rid of Delay For Expression (needed support from DAWs), increasing the number of instruments and ESPECIALLY number of articulations for each instrument...) A bit pricey for what it is, but it is definitely extremely flexible. Solo violins are tear-inducing. It just breathes, like samples will never be able to.

Can you see where these products will be in 5 years? My guess: we will probably have WIVI 2.5 or something like that, with all strings modeled, Synful will be completely done and it will sound exceptionally great, and Pianoteq is gonna be somewhere around version 5, with bigger user interface, different rooms to put the piano in, different materials of surrounding walls, different mic types, with glistening new models (definitely MORE COMPLEX ones, to awe all those that still claim Pianoteq sounds synthetic even after all those years of research!), definitely a couple of uprights, a Bosendorfer, and some more vintage keyboards (Pianet, RMI Electrapiano, a couple more harpsichords, Chromatic Percussion addon with vibes, celesta, marimba, glockenspiel, etc...)

Do you believe this all is possible and what are your opinions on this? I would also like to hear Julien, Philippe and Niclas on this topic


Peace out!

Hard work and guts!

Reply by Beto-Music

  • Avatar
  • Beto-Music
  • Member
  • Offline
  • Registered: 02-05-2007
  • Posts: 2,783

Re: OT: Reflection and discussion topic - in 5 years: modeling >> samples?

Some people say that in 10 years computer will use superconductors, and will be able to get 500 faster than today.


In 20 years maybe they get true artificial intelligence, and we will finally be able to say that there is inteligent life form on Earth.




But... 
Did you know that in the 60's many people said that today there would be families living on the moon?

Last edited by Beto-Music (13-02-2010 05:15)

Reply by Ecaroh

  • Avatar
  • Ecaroh
  • Member
  • Offline
  • From: Finland
  • Registered: 21-06-2009
  • Posts: 459

Re: OT: Reflection and discussion topic - in 5 years: modeling >> samples?

How much of it is about the model and how much about the computing power? If we had faster computers do we have a knowledge to use this power? Of course with increased power we could decrease latency which is very important in itself. Maybe it's just SO complicated thing to model real world... I cannot imagine how many physical things are there in moment when hammer strikes strings. (I've speculated that maybe this first moment is still the hardest thing to catch)

I remember somebody raised the issue about PTQ "supermodel" which could not (at the moment) play in the real time but calculate MIDI files to audio... So do we already have a better model if we don't need to worry about the computing power? In other words, is this supermodel already somewhere there waiting for better computers?

Last edited by Ecaroh (13-02-2010 09:42)

Ecaroh's Website

Reply by EvilDragon

  • Avatar
  • EvilDragon
  • Member
  • Offline
  • From: Croatia
  • Registered: 27-06-2008
  • Posts: 3,994

Re: OT: Reflection and discussion topic - in 5 years: modeling >> samples?

Ecaroh wrote:

I remember somebody raised the issue about PTQ "supermodel" which could not (at the moment) play in the real time but calculate MIDI files to audio... So do we already have a better model if we don't need to worry about the computing power? In other words, is this supermodel already somewhere there waiting for better computers?

I think this is the case. I remember Philippe talking in an interview about optimizing the complexity of his piano model, so it can be calculated in realtime. I guess he's doing a LOT of tricky choices to make CPU consumption relatively small.

Well, as CPU power progresses, he will not need to do that - we could have a checkbox in the Options "Enable complex model" or something like that - it would raise CPU cost, but I guess it would sound downright amazing

So yeah - the more complex models are already here. They are just "made to fit" into running on current hardware.

For example, the latest Pianoteq achieves 256 notes of polyphony with low latency on an i7 CPU, with laughable 5-7% of CPU usage! Imagine how Pianoteq would sound with a more complex model, that would take 20% on such a beast of a CPU!

Last edited by EvilDragon (13-02-2010 11:12)
Hard work and guts!

Reply by Ecaroh

  • Avatar
  • Ecaroh
  • Member
  • Offline
  • From: Finland
  • Registered: 21-06-2009
  • Posts: 459

Re: OT: Reflection and discussion topic - in 5 years: modeling >> samples?

Let suppose that it's there already. So why don't they release it? For many of us it would be just great to have this chance to make better piano tracks with it. You could use "normal model" to play and record and then "supermodel" to calculate audio tracks from it. For recording purposes it's the real output which matters most, no matter how it's done.

I think many people would put their money to this "supermodel" add-on. I still doubt that there's something already done that is clearly better, maybe I'm wrong... Only Modartt can answer that.

Ecaroh's Website

Reply by EvilDragon

  • Avatar
  • EvilDragon
  • Member
  • Offline
  • From: Croatia
  • Registered: 27-06-2008
  • Posts: 3,994

Re: OT: Reflection and discussion topic - in 5 years: modeling >> samples?

I think it's a part of their marketing plan

Sell it in steps For example, I think v4 will have the supermodel mode. It definitely SHOULD have it!

Time for a reply from Philippe or Julien

Last edited by EvilDragon (13-02-2010 11:25)
Hard work and guts!

Reply by julien

  • Avatar
  • julien
  • Admin
  • Offline
  • Registered: 19-04-2007
  • Posts: 1,026

Re: OT: Reflection and discussion topic - in 5 years: modeling >> samples?

There is no such super model. There is only our current model, which we try to refine and improve in the directions that matter (soundwise), while keeping aside all useless details (those that do not affect sound). This is were the things get difficult, figuring what is worth being refined in the model, what will bring improvements to the sound, and what is useless. Of course CPU considerations are also important when choosing whether to implement or not some feature, but so far it has not been such an issue that forced us to postpone an interesting feature.

julien's Website