Perfect piano?

Multi-platform modular music creation studio
Darkhog
Posts: 245
Joined: Wed Apr 06, 2011 11:03 pm

Perfect piano?

Post by Darkhog » Thu May 16, 2013 10:45 pm

Was anyone able to make perfect piano using synthesis, that sounds like real piano? I'm tired of so-called "synth piano" that doesn't sound like a piano at all, and using samples not only makes files bigger, but piano sounds bad at lower/higher pitches, because of the way SunVox and most other trackers do sound pitching - changing sample length along with sample's tone. There are libraries that can get you around this (SoundTouch, anyone?) as Audacity has shown, but still, it isn't used in sunvox.


So was anyone able to make perfect piano using only generators? I had mild success with MultiSynth and two analog generators, but it still doesn't sound like real piano:
experiment1.sunvox
(9.84 KiB) Downloaded 382 times
Had someone better luck?

organic io
Posts: 439
Joined: Fri Jun 24, 2011 12:31 am
Location: USA - North Carolina
Contact:

Re: Perfect piano?

Post by organic io » Thu May 16, 2013 11:07 pm

To get a real piano sound, I would just try to find an XI instrument that is multisampled from a real piano. It would have at minimum one sample per octave to sound OK, and preferably a separate sample for each key :)

User avatar
samrai katt kovboy
Posts: 308
Joined: Wed Jun 13, 2012 11:07 am
Location: Spain
Contact:

Re: Perfect piano?

Post by samrai katt kovboy » Thu May 16, 2013 11:15 pm

+1
SOOOOOOOOOOOOOOO MANY CIRCLES

gigi
Posts: 91
Joined: Sat Dec 15, 2012 3:33 pm

Re: Perfect piano?

Post by gigi » Fri May 17, 2013 12:07 am

unfortunately, perfect acoustic piano will never be posible to make using actual generators, because they generates harmonic timbres, or an acoustic piano has a inharmonic timbre.

the difference between, is that the "harmonics" of a real piano note are not perfect tuned acording to the base freq of that note, but higher - or sv generators generate perfect tuned harmonics for a given base freq.

so we need an additive generator or a generator with a parameter "harmonicity".
harmonics are natural numbers multiple of the base freq. Example : if base freq = 100 hz => harmonix are 200, 300, 400, 500 etc hz, so 2x, 3x, 4x etc (x= base freq).

"harmonicity" parameter have to let us set an arbitrary non natural, but real number value. For example : harmonicity range may be real numbers from 1 to 32. In this case, the harmonic formula will be N x BaseFreq x Harmonicity.

For real pianos, we can try a value from 1.01 to 1.2 for the harmonicity.

I wrote this message very quick. I hope I was clear. If not, let me know.

Darkhog
Posts: 245
Joined: Wed Apr 06, 2011 11:03 pm

Re: Perfect piano?

Post by Darkhog » Fri May 17, 2013 12:52 am

I think it was clear, but couldn't we just use several Analog Generators with different Frequency2 parameters, then play with wave type, volume and attack/decay times? Then to keep those synths in sync, we would pass notes via Multisynth.

Certainly we would more than two AnaGens in my example posted above, but with enough time we would be able to develop chain so close to real piano that is almost indistinguishable (certain ZynAddSubFx synths are really close to piano, but I think we could get closer) and then envelop it into meta module.

Could anyone try it and see if he/she can come up with interesting results? I cannot be bothered with it, because I am here to make music and since I couldn't get working piano in realistic time frame... Anyway you can use my file (attached in OP) as base.

@organic.io the problem is that it would increase size of music file. I want realistic piano in my 120k game (self-imposed goal, not for any compo or something).

gigi
Posts: 91
Joined: Sat Dec 15, 2012 3:33 pm

Re: Perfect piano?

Post by gigi » Fri May 17, 2013 2:23 am

Darkhog, it is possible to use several anaGens to make it and detune each anaGen higher and higher, BUT :

1. in a real piano, the harmonicity of each string is different, because it depends on its elasicity, and the elasticity depends on the lenght and thin of each string. The formula is very complicated, but the result is that the harmonicity is higher for the high harmonics and for the high notes. For this reason we cannot use a single sample or a aingle cup of settings in a metaAnaGens for all the notes of a piano.

2. for each note in a piano there are several strings, very little detuned.

3. a real piano has a lot of noises. The hammer noise on attack,for example, is very specific and we have to reproduce it too. In the acoustic laboratory in paris where i worked, a real piano recorded, got the hammer sound erased with a special scientific software (audiosculpt on mac) and guess the result : it sounded very electronical !! So, you see, noises are very importand in acoustic timbres.

This makes the unicity and beauty of a real piano sound and the insane task for us to reproduce it.

I tell you this ideal sound just to know the ideal direction to go.

So we'll never get a perfect sound, but try to get closer and closer. The problem is that for a real time piano 100% synthetised, the cpu will be over charged and such a project will be useless.

I will try to get a good compromise and I'll tell you the result.

gigi
Posts: 91
Joined: Sat Dec 15, 2012 3:33 pm

Re: Perfect piano?

Post by gigi » Fri May 17, 2013 2:26 am

sorry, not the harmonicity, but the INHARMONICITY is higher for the high harmonics and high notes.

gigi
Posts: 91
Joined: Sat Dec 15, 2012 3:33 pm

Re: Perfect piano?

Post by gigi » Fri May 17, 2013 2:39 am

in the same acoustic laboratory Pianoteq was born by one of my colleagues. Pianoteq is a 100% synthetised piano software. As I know, it needs off line rendering to bring us the sound, just because of the quantity of operations needed for a natural beautiful sound.

gilzad
Posts: 92
Joined: Wed Dec 19, 2012 6:46 am

Re: Perfect piano?

Post by gilzad » Fri May 17, 2013 10:51 pm

Interesting discussion, guys!

A week ago I just started building my synthesized piano. Exactly as described here, with the intention to sound natural and not like a DX7 or Rhodes.

My approach is total by-ear, me being the only one to decide when it sounds convincing enough. I was gonna mention Pianoteq but I'm even luckier to see that you're involved in the same lab, gigi! Excited to learn from your knowledge.

Once in a demoparty some group has shown a demo that created very authentic sounds based on FM-synthesis. I kinda went a mixture-approach, combining FM-synthesis with additive synthesis, applying different envelopes. I notice that it sounds quite convincing in the upper octaves but very synthish in the bass range. I got it solved by adding formants to the bass-range (e.g. by adding 5ths, like Darkhog), which of course ruins the upper octaves at first. So I'm gonna create a split keyboard with a bunch of Multisynths. A detailed EQing is inside, too (you'll get the desired 10-Band-EQ with that MetaModule then). But with all the combinations and DSPs gigi's hint will become true. The CPU-usage won't be very low. Maybe we can try to get a good sound first, then monitor the CPU usage and see how we can reduce that.

I just need some extra time, because I updated my OS, which always comes with a bit of headache. And I'm very excited to see your versions, gigi!

User avatar
samrai katt kovboy
Posts: 308
Joined: Wed Jun 13, 2012 11:07 am
Location: Spain
Contact:

Re: Perfect piano?

Post by samrai katt kovboy » Sat May 18, 2013 12:09 am

THIS IS REALLY EXCITING!
Looking forward to what the outcome will be!!!
SOOOOOOOOOOOOOOO MANY CIRCLES

gigi
Posts: 91
Joined: Sat Dec 15, 2012 3:33 pm

Re: Perfect piano?

Post by gigi » Sat May 18, 2013 1:40 am

even if any sound engineer know it, I want to mention that the quality of a poory synthetised piano sound is very depending on the mix.

even a bad one will sound allmost great if the mix covers it intelligently.

so is not always necesary to create a super real piano sound if the mix context of the song can help us by offering ways to cover it intelligently.

a way to try it, is to create the others intruments first and then synthetise a piano that just fill the gaps, until it sounds credible. Once done, if we solo that piano, we'll see that it's far to be a nice one (in solo), but who cares ?

this is the most economic way - I'm sure.
What do you think, guys ?

Darkhog
Posts: 245
Joined: Wed Apr 06, 2011 11:03 pm

Re: Perfect piano?

Post by Darkhog » Sun May 19, 2013 5:51 am

The problem is, aside of drums piano is the only instrument in song I plan to make.

gigi
Posts: 91
Joined: Sat Dec 15, 2012 3:33 pm

Re: Perfect piano?

Post by gigi » Sun May 19, 2013 1:53 pm

oh, in this case, yes, it have to be credible even in solo mode. :)

jcage
Posts: 45
Joined: Fri Jan 18, 2013 1:25 am

Re: Perfect piano?

Post by jcage » Sun May 19, 2013 6:39 pm

Being a sampler freak I created a 4-layered stereo piano as metamodule. 134MB... to big? Not for what it sounds like. Fazioli grand piano, sampled by folks who know their business, provided by palmitos some time ago. Took me some time to translate..

User avatar
samrai katt kovboy
Posts: 308
Joined: Wed Jun 13, 2012 11:07 am
Location: Spain
Contact:

Re: Perfect piano?

Post by samrai katt kovboy » Sun May 19, 2013 9:18 pm

IS there a possibility to download this from you?
Please...
SOOOOOOOOOOOOOOO MANY CIRCLES

Post Reply