I came across this suggestion in another forum, and I really like it.
Pianoteq's modeling is very simplified due to current processing power constraints. But what if we could use the current Pianoteq models for real-time MIDI recording, and have Pianoteq render the sound overnight (for example) using a more rigorous model?
We've been speaking about that here also:
http://www.forum-pianoteq.com/viewtopic.php?id=1165
According to Modartt, the current model is best they have and there's no "super model" waiting for better computers.
Wouldn't it be interesting if some company had some classic pianos hooked up to a sophisticated player mechanism that would mechanically reproduce all the standard MIDI controller data for a MIDI performance. You upload your MIDI sequence. The piano is then recorded with an optimal mic setup and the resultant audio file is then downloadable. This could actually be done automatically, albeit, humans would be necessary to keep the piano in top notch condition and possibly alter microphone placement (though that could be done with robotics). You might have to wait your turn if it was a busy enterprise, but it is easily within the realm of current technology.
Cellomangler,
Such services already exist.....almost.
The ones I've seen still have a human to oversee that the recording is made properly, as far as I know.
Greg.
I'd also like for Modartt to rigorously increase the complexity of their model for version 4, and then offer it as an option in Options. It should be REALTIME, but more CPU demanding.
So, if you turn it off, it's back to the more simplified model, which is good for laptops etc.
I mean, with the way CPUs are advancing into hyperthreading and multicoring, it would be a SHAME if Modartt wouldn't take advantage of these new i7 (and later this year i9!) CPUs!
Why not a model that compute a library of samples to be used for the piano model, so that more processing time is left for the software.
I have read on another forum that samples are necessary in any case as for hammer noise and action noise, if it is the case more elements could be generated and a library build ?
A preview feature allowing to test the model before computing all.
Thats easy to ask !!
In my experience it's problematic to change piano sounds (export MIDI files to other piano sound/software/hardware). Reason is simply this: player will adapt his playing (more or less conciously) to MIDI controller & sound device which he's using. It's never the same when same midi-information is played through other system. Of course you can try to edit it to match it better but still it's very difficult. That's why I'm not interested in idea of sending my MIDIs to some real piano midi-player. Sound might be good but I'm quite sure that it's not expressing my musical ideas and nuances correctly.
Many sample libraries have this problematic even when you're playing/recording with them in real time. Those softwares like Ivory doesn't give you good response like Pianoteq. Individual (sampled) sounds can sound great - And of course they do; they are well recorded individual sounds!
I like the idea of complicating PTQ model if Modartt has knowledge to do this to make it sound better (in faster computers or rendering MIDIs with slower machines). I believe that with this "two models model" - one for playing/recording and other for real good piano tracks - the play-response -problem is not there. But: as said earlier, Modartt hasn't this "super model" (yet).
I also believe that attack is very difficult to model; so many physical things / variables are there in the very first moment. Still I wouldn't give up for using samples. In fact in my opinion Roland's V-piano has succeeded quite well to make this punchy, woody attack (maybe they're using samples and just telling lies about their model ...)
In my experience it's problematic to change piano sounds (export MIDI files to other piano sound/software/hardware). Reason is simply this: player will adapt his playing (more or less consciously) to MIDI controller & sound device which he's using. It's never the same when same midi-information is played through other system. Of course you try to edit it but still it's very difficult.
Many sample libraries have this problem even when you're playing/recording with them in real time. Those softwares like Ivory doesn't give you good response like Pianoteq. Individual (sampled) sounds can sound great - And of course they do; they are well recorded individual sounds!
"Export midi files to other piano sound/software/hardware".
This is absolutely correct, and it seems we keep forgetting it. I'm certainly not a great pianist, but whenever I'm at a different piano, it takes me a while to adjust to the new instrument (the better the pianist, the less time it takes).
Samples - they are after all just samples and as such are only a small part of the overall sound. Samples keep trying, but truthfully they will never really succeed because the nuances of piano sound (which are much more than one note) cannot be sampled.
I'm starting to think that people that use and like samples, may not actually play the piano, or have never played a good acoustic grand piano. Whenever I play my DP for quite a while, and become accustomed to it and then sit at a grand, it's always a surprise and revelation at how different they are.
Glenn
Cellomangler: The Yamaha Disklavier can do just that (under some human supervision), if I understand your musing. As others have said though, every controller will have its own velocity and timing response. Disklavier works best when MIDI sequences are recorded on a Disklavier. (I was commissioned to write a piece for it several years ago and messed with it quite a bit.
Glenn:
I agree and am one of those : ) I'm less than a hacker as a pianist, although I have access to a well-maintained Yamaha C7 concert grand. I do love to PLAY Pianoteq, and the synthesist and engineer in me loves the flexibility. Modeling is the future of digital instruments and Pianoteq is a musical instrument today. However, at this point, to create recordings of my projects, I decided to use Garritan Steinway. It sounds more like the acoustic piano in my ear because it is. Again, Garritan Steinway isn't an instrument in my view, but is very usable recordings of an instrument, done in a limiting, but excellent hall.
Among the problems, of course, is the limited # of static samples for each key. (I won't go into resonance, miking, and other issues right here.) Garritan does a fine job of blending, but many breaks between the 3-6 velocity layers are obvious and audible, particularly those at the extremes: velocities around 13-14, and 98-100. This means that after humanizing velocities on sequences for Pianoteq, I had to re-tweak almost all notes for G.Steinway to keep passages from jumping in and out.
So... When playing live for enjoyment, composing, or recording sequences in real time, I'll always use Pianoteq. Same for sound design, the other modeled instruments, rock recordings, etc. As an INSTRUMENT, there is no comparison for me. Then, for concert recordings featuring the piano in a (simulated) live space similar to Troy Hall, in conjunction with Altiverb to match other instruments to that space, I'll edit most notes of the recorded sequence to best fit the Steinway and use that, because at this moment in time, as RECORDED SOUND, there is no comparison for me, no matter how much I work with Pianoteq settings.
As another approach, the new Vienna Symphony Lib Imperial takes a shock and awe approach to sampling (1200 samples per note, 550 gigs, $1000. or whatever), but even if I could justify the expense, I just don't really love the sound that results from it all.
Sorry if this got off-topic a bit, but I was inspired to respond based on what I am working on at the moment.
Steve
I agree with many points of Steve. In fact I'm in a situation of recording some serious piano stuff and I'm still thinking about different digital solutions (acoustic piano isn't a possible this time). Musically I'm searching for expressive and clear studio-like piano sound for jazz group. As been said, Pianoteq is one of the best when you think about playability and response. I have argued elsewhere that still it has some problematic especially in lower part of piano.
So far I've tested some sampled alternatives like Ivory and then some newest hardware like V-piano or Yamaha S70SX. From all those maybe V-piano is closest thing what I'm after for - it has a marvelous keyboard and the sound is quite even (no better or worse areas). Unfortunately it's too expensive and not so usable if you think about taking it with you to some gig. For recording purposes V-piano has limits too: it seems to me that it's always more or less players piano; you cannot "record" it from distance like PTQ.
About sampling technology: In my experience amount of sampling - velocity layers etc. - doesn't guarantee better sound! In fact it can be quite opposite. More samples you have more there's a chance that those sampled sounds aren't equal. I can hear in Ivory that some keys doesn't respond or sound like keys near to them. For a player this isn't good thing. Maybe it's because of the recording device (machine that hits keys) or room acoustics (some notes sounded different in their recording room), I don't know.
And finally about technological development: I can clearly see that many companies are putting their effort to combine sampling and modeling. Maybe it's because of the difficulty of modeling the attack? I don't know how do they actually do it but the main goal is similar with modeling: give a player chance to control many things, resonance functions, no looping and no audible velocity layers. Let's see how do these new piano sounds manage? I'm waiting for checking new "Super Natural" expansion for my Roland RD700gx. This is just one example of the new hybrid technology.
I'm waiting for checking new "Super Natural" expansion for my Roland RD700gx. This is just one example of the new hybrid technology.
I just installed the Super Natural expansion this weekend. If you liked the playability of the existing piano you'll definitively like this one. The piano sound itself has much more warmth and body. It also comes with some nice upright and ragtime sound.
But.........despite the fact that Pianoteq still lacks a bit of the real piano sound (in the higher regions there's still a little "ting"), the sound of Pianoteq is more clear and less (or completely not) nasal; I still notice some "nasal" sound in the mid-region with most hardware (kawaii, roland, yamaha, .... had all of them).
And something I'm missing from my Roland RD700GX is the ease to change to different tunings. But the keyaction is the best I've felt (the same one used in the V-piano) but that's personal ;-)
I just installed the Super Natural expansion this weekend. If you liked the playability of the existing piano you'll definitively like this one. The piano sound itself has much more warmth and body. It also comes with some nice upright and ragtime sound.
Just curious: what kind of parameters do you have with new expansion? Do you think you can really make different pianos with it?
Just curious: what kind of parameters do you have with new expansion? Do you think you can really make different pianos with it?
I have to dive deeper into this expansion board but..... at first sight it's just new sounds with the possibility to change parameters which are the same as with the default sounds. As far as I know: they took the sounds generated by the v-piano and put them onto the expansion board so this is fixed, so no: you can't create new sounds. Also, there should be a different engine behind it (for resonance, duplex etc.) but actually I couldn't discover any change in playability. It sounds a bit negative...but...the sounds are better than the default sounds. Because of the fact that the sound experience is a matter of taste, each person should decide for themselves whether it is worth the extra money. For me it is.
Ecaroh wrote:Just curious: what kind of parameters do you have with new expansion? Do you think you can really make different pianos with it?
I have to dive deeper into this expansion board but..... at first sight it's just new sounds with the possibility to change parameters which are the same as with the default sounds. As far as I know: they took the sounds generated by the v-piano and put them onto the expansion board so this is fixed, so no: you can't create new sounds. Also, there should be a different engine behind it (for resonance, duplex etc.) but actually I couldn't discover any change in playability. It sounds a bit negative...but...the sounds are better than the default sounds. Because of the fact that the sound experience is a matter of taste, each person should decide for themselves whether it is worth the extra money. For me it is.
Let's see, I'm waiting for my expansion board that's is coming in few days, I hope. If it's like Hvaartsen said, we're dealing with new kind of idea which is in fact close to first topic of these posts: using no-real-time-modeling for making sounds.
Let's speculate it further: Roland is using complex (very CPU demanding) model of V-piano and they've rendered sounds with it (maybe mostly attack sounds) and now they put them together in their Supernatural Board. Is this method good or bad? Maybe it's the future way, of course if their model works well. At least there're some benefits compared to sampling or complete modeling. If memory is cheaper than putting fast CPU to hardware maybe this is reason number 1. Reason number 2 is obviously the fact that with a good model you can have very easily enormous amount of different pianos in different spaces; Imagine that you had to record all these. With "no-realtime-modeling" you can have as many layers you want and sounds are even; in other words there's no problematic I have noticed with gigalevel sample pianos: keys sound and respond unevenly.
But (and of course this is a big BUT): there's no interaction (resonance) like in real-time modeling. Maybe there's a chance to model this part and use kind of hybrid technology, I don't know. Anyway with these visions now we have in fact at least four different possibilities to make piano sounds: 1. Old fashion sampling; 2. real-time modeling; 3. No-real-time-modeling (=using models to create wavs or some audiopianosamples) 4. hybrid technologies which combine 1+2, 1+3 or 2+3.
Finally there're many reasons to use one or other method. One, I think that is quite important here at the PTQ forum, is kind of philosophy: making complete (and hopefully real-time) model is intellectual challenge - kind of very old fantasy of beating the nature (using sampling feels like compromise and turning back)...
I was of the understanding that SuperNatural was sample based. Certainly, it sounds too good for a V-Piano IMHO.
I'm not sure what aspects of the V-Piano have been used, though - I'm very interested to know this.
Greg.
I was of the understanding that SuperNatural was sample based. Certainly, it sounds too good for a V-Piano IMHO.
I'm not sure what aspects of the V-Piano have been used, though - I'm very interested to know this.
Greg.
I installed my Supernatural Expansion yesterday and I must say I'm quite impressed: with this board you can get much more from the good old RD700GX. It's hard to know how they did it (combining multisampling with V-piano technology) but it's very expressive and powerful piano sound and still providing also soft and singing qualities. Sound is also much more even in all areas (I didn't like old RD's upper register for example). I cannot hear any velocity layers, so maybe it's like they promised.
In my opinion with this expansion RD comes surprisingly close to their V-piano. Compared to Ivory Steinway Roland's new sound comes also closer to that and it's much more playable and natural responding than Ivory.