|
Post by pabstblueribbon on Dec 23, 2018 21:33:50 GMT
I own an oldschool sampler from E-MU (the ESI 2000). I would like to use some piano sound libraries for it. The problem is I only have 128 Mb, while many present piano libraries are much larger in file size. Also there's the thing it only supports two-layer presets. I can solve it to create multiple presets each containing a specific velocity range. I can use my stagepiano as master keyboard which can play 4 presets simultaniously. This will give me to use 8 velocity layers in total.
It's quite amazing how my stagepiano Roland RD700sx containing for about 64 Mb of piano samples can sound much more expressive and realistic than many software pianos with multiple gigs of samples. My Roland stagepiano contains samples of each key in multiple layers. To squeeze that into 64 MB the samples have to contain loops which you'll hear after about 8 seconds. Creating loops in piano samples is very challenging and often lead to terrible results. I've noticed very short loops usually give better results than longer ones.
I've got an idea for more appealing sounding results:
This picture represents the sound wave of a single piano tone in its low register. The first part lasts for about 8 seconds. Then the loop appears: a very short loop so won't hear anything repeating. The loop will sound very flat, comparable with an unedited sine wave.
The loop part will be edited into audio editor software. First step is to repeat it so many times the audio fragment lasts for a couple of seconds. Then it can be edited to simulate the vibrations in the piano string. It's nice to listen to the diminishing sound of a piano after you've pressed the keys, because then you will hear a kind of pattern repeating over and over until the sound is silent. You'll hear that same kind of vibrations in other stringed music instruments like a guitar or in some percussion instruments like cymballs. I assume you need to know some basic knowledge about sound acoustics to understand what's happening. I've read something about it in the far far past, but it ain't 100% clear to me what actually happens. So if you can explain it to me or tell what I should study to understand it you will help me a lot. The meaning is to simulate this for the loop part of the piano sample. It doesn't have to be 100% realistic, but it has to be appealing to listen to. Because then loops can be used in the samples of a piano library.
By the way: in the picture you see a gradient transitions between the first part and the looped part. This is to represent a crossover transition when attaching the looped part to the first part: maybe this will lead to good results.
It's a very challenging idea and perhabs very hard or not feasable at all. So I would like to get feedback about this!
Thanks in advantage,
Michael
|
|
|
Post by rock on Dec 23, 2018 23:00:20 GMT
Hi, Sorry I did not read your entire post but IMHO, if you want great digital piano sounds try this one: www.pianoteq.com/pianoteq6I have only downloaded the demo version so far but it sounds great. (BTW, I don't do serious recording or production and mostly play live. When I do record, I play in real time.) To run pianoteq, you do need a modern proc like and i5 or i7. You don't need tons of memory because it's a modeler NOT a sampler. There are NO samples. The Stage version is the most basic and the Standard has more tweeking capability while the Pro is uber teekable even to the point where you can tune (or de-tune) each of the two or three strings in each unison besides many other tonal aspects. You can download FREE demos for the Stage and Standard. For my purposes, I found the Stage version (simplest and cheapest) flexible enough but you can always upgrade so you can't lose. As I mentioned, I have only down loaded the demo so far but the demos are fun to play with. The demo has a bunch of high and low sharps disabled and shuts down after 20 min. requiring a restart of the program but I think you can get a very good idea of what your'e getting into and have some fun playing it at the same time. If you want to leave the fooling around with samples behind, try modeling. If you buy this, tell modart to send me a check Cheers, Rock
|
|
|
Post by Hexspa on Dec 24, 2018 3:07:15 GMT
Ok, so you want a sample to fade naturally, right? You need to set the amplitude envelope sustain level to zero, the decay portion to be realistically long, the release setting to be appropriate for your played part and you need to loop the part of the sample that is correct for the part also. It comes down to how long you have the keys depressed; long-held notes will probably need more of the middle of the sample looped whereas shorter notes will want the tail end so you can get a bit of a decayed sound. With shorter notes, the release becomes more important too since you'll be relying on it to provide the 'in-between' ambience.
Couple that with a decent reverb or delay and you should be good to go.
|
|
|
Post by pabstblueribbon on Dec 24, 2018 22:15:43 GMT
@ Rock. Thanks. I will check it out. If it comes to piano libraries I've recently found two other good ones as donationware: ivyaudio.com/www.productionvoices.com/estate-grand-le-for-sforzando/I haven't tried the first one yet. It's a couple of Gigabytes. I've noticed the samples are 96 kHz by 24bit. Personally I think it's a good idea to reduce it to 44.1 kHz by 16 bit, because otherwise it will take a lot of RAM from your system. The one from ProductionVoices sounds really good. I have to mention though the piano sound of my Roland with only 64 Mb sounds just as good. There's a myth the bigger the sound libraries are the better the quality should be. That's definitely not true. From the website of Ethan Winer I've discovered a soundfont of a cello which is just a few megabytes but it sounds way better than the majority of other cello SFZ-files and soundfonts which can be hundreds of megabytes. Coming back to piano sound libraries: I think it's more important to know the behaviour of the piano sound instead of just capturing as much samples as possible for a huge sample library. I prefer the ones which may be slightly less realistic due to less samples but are programmed for expressive playing. When you press the keys on the piano the volume decreases of course, but also its brightness and duration and probably other features as well. I imagine when you want to program such decreases in a soundfont, Kontakt library or whatever they ain't lineair but in curves. (I don't know how to explain it in words; I hope you understand me.) Also I think it differs by each string.
|
|
|
Post by pabstblueribbon on Dec 24, 2018 22:22:41 GMT
Hexspa Thanks for replying. These are very usefull tips! In my former post I've replied partly to you as well.
|
|
|
Post by Hexspa on Dec 24, 2018 23:51:44 GMT
Hexspa Thanks for replying. These are very usefull tips! In my former post I've replied partly to you as well. As far as the darkening-over-time, that's what a filter envelope is for. Not sure what kind of parameters you have available in your machine but Massive, for instance, has freely-assignable envelopes as well as filter keytracking. The latter feature allows you to adjust the filter cutoff depending on which part of the keyboard you're playing on. I know it has a few breakpoints but you might even be able to add more to adjust it on a per-note basis. Way too much tweaking for me but the possibilities exist.
Here's a video you might find relevant where Espen Kraft does some sampling to include amp envelope adjustments.
|
|
|
Post by rock on Dec 25, 2018 4:12:09 GMT
Hey PBR,
Thanks for reading my reply and I'm sorry I didn't address your question but I don't know much about woking with samples. BUT... from your reply it seems you didn't get my point. (Maybe you did but it doesn't sound like it. Sorry if I misunderstood.)
Here it is: Modeling software does not use samples! Modeling software creates sounds in realtime from... drum roll please...MODELS. Models are mathematical equations based on the physics that characterize all the ways strings vibrate and are influenced by all the various parts found in a piano. The cool thing about this is that, since there are no discrete samples involved, extremely small changes in midi input will influence the output in subtle ways without the need to have tons of samples crossing over from one to another and all the other tweaking that goes along with it. It's a whole new ball game!
And as I mentioned, there are NO samples so very little memory is needed compared to a sampler. On the other hand, the processing speed and capability of a somewhat modern CPU is required to run these models in real time. Having said that, I can easily run this kind of SW with a 2012 mac mini 2.6GHz i7. I have 16 GB ram but you don't need much because there are no samples so the SW is fairly compact.
I don't have anything against samples but if your goal is a realistic piano you've got to check out this pianoteq. Did I mention the demo is FREE?
|
|
|
Post by pabstblueribbon on Dec 25, 2018 22:21:28 GMT
Hexspa I will use Polyphone for programming the soundfonts. I recently started to work with it; it's such a gem because it contains a couple of very handy features. It can detect loops automaticly, manually setting loop points is easy as well, to set the range of the samples you can make use of your mouse by dragging the desired area in a graphic window. I have little experience with using EQs. Maybe this software can be handy for learning. I mentioned before I've found piano libraries with many velocity layers. Perhabs it can be useful to compare the frequency spectrum of the samples to find out how the sound changes with each velocity. But first I will watch your video. My own ideas are very time consuming I suppose hahaha. Programming piano soundfonts can take months maybe years.
|
|
|
Post by pabstblueribbon on Dec 25, 2018 22:22:25 GMT
rock I'm sorry, I haven't understood your question very good then. It's definitely worth checking out. Maybe it can be of use as well when getting to know what I want to understand
|
|
|
Post by rock on Dec 25, 2018 23:18:39 GMT
Hi PBR,
I think you are right that you didn't understand but I'm sorry, I didn't even ask a question. What I'm doing is trying to provide you with an alternative way to play a digital piano.
I can tell you are interested in working with samples and looping, manipulating and all that, that's cool. But I'm simply telling you that there is, IMHO, a superior way to get a great piano instrument using a completely different technology which is modeling.
If you go with the SW I suggested, you will leave behind any cares of detecting loops and all that, this SW is plug and play. There are user controls but they are directly related the sound quality of the modeled components like string length ,sympathetic resonance and duplex scale; sound board impedance, cutoff and Q; damper position and duration; tuning unison width and octave stretch; hammer hardness and on and on.
Sampling is still of course widely used but it's old school compared to modeling. If you like to get "under the hood" maybe sampling will allow you more ability to mess with the "nuts and bolts". It's of course all up to you what's most important.
Cheers, Rock
|
|
|
Post by pabstblueribbon on Dec 26, 2018 14:25:27 GMT
I must apologise too, because I have a very short attention span when reading. If I was able to focus better than I would have studied everything about sound acoustics and sound synthesis. I wanna know everything at once and of course that's not possible. I'm already glad I recieved useful replies here ~~~~~~ First to go back to my question. Maybe I've found a VST plugin which does what I want: www.audiopluginsforfree.com/algorbloom/ I haven't tried it yet though. ~~~~~~ I've heard of physical modelling. Many years ago I encountered a very expensive digital piano from Roland which produced its sounds with physical modelling. If I get it right the present generation of digital pianos make use of both sampling and physical modelling. I'm interested in sampling because it's a very creative proces. The old standard SF2-format offers so much possibilities already for modelling the sound you can design synthesizer sounds as well from samples of simple sine waves or saw waves etc. SF2-files can be converted to other formats for oldschool samplers, so you can program your samples on the computer and later export them to the machine. For me personally that's interesting because oldschool samples have their own character which can be very. Also I found out several music creation apps for Android make use of SF2-files; you can replace those with your own SF2-files. I find it very interesting to know this. However I must admit I have to be aware not to spend so much time with it I'll keep delaying making music hahaha
|
|
|
Post by rock on Dec 27, 2018 1:01:20 GMT
OK that all makes sense. The pianoteq SW (much like the Roland V-Piano) is mainly for reproducing instrument sounds and even though you can modify and customize, it's primarily for playing real sounding pianos etc. So if you want to create your own unique sounds, the techniques you're doing is a good way to go.
The main reason I took us down this path is that it seemed to me that you were "bending over backwards" trying to get good piano sounds, but I now see you are as much interested in unique sound creation so Go-for-it!
Cheers, Rock
|
|
|
Post by Hexspa on Dec 27, 2018 1:19:55 GMT
sound acoustics and sound synthesis. Here on this forum you can find just about all you need to know about practical small room acoustics. I can recommend Syntorial for learning synthesis.
|
|
|
Post by Ethan Winer on Jan 2, 2019 16:15:02 GMT
|
|