I Found a free CFX soundfont and it works great on my PSR S970

[BogdanH]

Sorry to hear that Mark.. no idea why's that.. maybe file got corrupted somehow.
Right now I've sent you pack again.

Bogdan

[Amwilburn]

Yup, it was a corrupted download, this time it worked!

Now to install them to give them a whirl!

Mark

[Amwilburn]

Hmm... the CFX and the Kawai sound exactly the same ;p The Steinway (which is the best piano in the world) sounds the worst from this soundfont.

The Yamaha, Kawai, and Fazioli all fared better on this pack; I think because they sounded less 'plunky'. The Bosendorfer, normally a very gentle sounding piano, sounds super thin and plunky; like all notes were recorded at fff.

Still, for the 64mb sample sizes, I think the Yamaha and Faz are worth installing! Thanks so much for doing this

Mark

[BogdanH]

Thank you Mark, for sharing your opinion.

From what I've research so far, all soundfont samples I could find for free are fff only. One can buy samples that also include other velocities, but they cost a fortune.
And then... if voice that only contains fff samples has size of about 40MB, then voice having five velocities samples, would have size of 200MB -is still acceptable for hi-quality voice, though. Anyway, to keep voice size small, lower velocities are usually added as "artificial" layers, where velocity is only simulated (by frequencies cut-off, etc.).
But I didn't do that in pack that I've sent you, because I wanted you to hear samples as they are -and also because I wasn't sure at that time, if it's worth bothering.
Btw. I'm quite sure, that Lyzen didn't actually record samples (he only made a collection of samples that he found elsewhere). I say that, because I found exactly the same CFX soundfont which additionally contains higher octave samples. That's to be expected, because not everyone is in the position to record $200.000 piano.

Almost forgot.. later I found out, that most of this samples have L/R channel reversed. Means, when you play low key, it will sound louder on right speaker and vice verse. I think that's not because there's something wrong with samples -it might depend on where microphone was positioned at recording. But that can easily be solved when creating voice.
I also realized, that some (not many) notes in samples contain short high pitch tone at the beginning when hammer hits (it's most noticeable at low keys). In that case, one needs to choose next key sample (i.e. C#1 instead of C1). And that can additionally make a lot of work, because we usually notice these things after we install and play the voice.

It's interesting that CFX and Kawai sound the same to you... I can't confirm that.
Anyway, for me personally, CFX sounds the best (is most "pleasant" among them) and so I will try to make a "final" voice out of it.

Bogdan

[Amwilburn]

That's the thing, in real life, all 5 of these sound quite different from each other! THe CFX is the closest to how it should sound, the Kawaii sounding like a Yamaha is rather odd! And the Bosendorfer, normally a very gentle tone, sounds the most removed from reality!

[BogdanH]

Let me recap what I've learned so far...
First and most important: when we get some voice or soundfont (for cheap), usually we know nothing about it's source (the origin of samples). Ok, they say "recorded with hi-quality equipment" etc., but reality is usually far from that. Quite possible samples come from some (piano) keyboard. I mean, it's easy to get samples from keyboard we own, right? And then, these samples go trough many hands, where everyone tries to "optimize" them (which were already "optimized" by someone else before). Or someone just makes a little equalization and adds/removes reverb and calls it Steinberg -you get the point.
After saying all that, I would actually be very happy if I could get untouched samples from high quality digital piano (that has CFX voice built-in).

@Mark
I "studied" samples (of voices in pack that I sent you) and came to conclusion, that all voices (except CFX) simply have too short (too steep) decay. That is, after pressing the key, loudness falls too rapidly and continues with sustain. Turn off touch sensitivity, turn on sustain, hit C3 key (for example) and compare CFX with other voices. I don't say CFX is "great", but is still quite usable I think.
So I made a new CFX voice (because L/R channel needed to be swapped anyway) and used samples from another soundfont (which is less "optimized" I believe). Finally I deleted all other voices.

Thank you all for patience reading this 
Bogdan

[overover]

Hello everyone,

if someone doesn't know this site yet, you can find there a lot of high-quality free SF2 soundfonts, including many high-quality pianos, and in many quality levels / file sizes.
>>> https://sites.google.com/site/soundfonts4u/


Best regards,
Chris

[BogdanH]

hi Chris,
Thank you for pointing on this. Actually I downloaded few soundfiles from this site, but I just didn't have time to try them.

Bogdan

[Amwilburn]

Thanks Chris!

I just tried a bunch of them, shockingly unusable because there's a weird 'wash' sound as you sustain. Imagine if someone sampled white noise as piano resonnance. But I'll keep looking! Just be warned, none of the 237mb and below samples works great (the 'wash' is a lot less noticeable on the the C5 v2, but it's sitll there).

Maybe it's just an issue importing the sf2 directly into yem?

[overover]

Thanks Chris!

I just tried a bunch of them, shockingly unusable because there's a weird 'wash' sound as you sustain. Imagine if someone sampled white noise as piano resonnance. But I'll keep looking! Just be warned, none of the 237mb and below samples works great (the 'wash' is a lot less noticeable on the the C5 v2, but it's sitll there).

Maybe it's just an issue importing the sf2 directly into yem?

Hi Mark,

A long time ago I experimented with older versions of "Yamaha C5 Salamander Grand" and some "Steinway" soundfonts from the Soundfonts4U site. They actually sounded really good. After importing into the YEM, however, the reverb settings should be checked and optimized. (Often too much reverb is set.)

By the way, for quick viewing of sf2 soundfont files (to check the layers and samples inside) I use "Awave Studio". The trial version can be downloaded here:
>>> https://www.fmjsoft.com/awavestudio.html#main

You can also install the Coolsoft "VirtualMIDISynth" and load the desired soundfont there. You can set the VirtualMIDISynth as the MIDI output device of your DAW or a MIDI file player (I use "vanBasco Player"). However, the VirtualMIDISynth cannot be played in real time, i.e. you will have an audible latency (delay) if it is controlled directly by a MIDI keyboard.
>>> https://coolsoft.altervista.org/en/virtualmidisynth

To edit sf2 files I have installed the two free programs "Polyphone" and "Viena64", but I hardly do anything in this direction at the moment.
>>> https://www.polyphone-soundfonts.com/
>>> http://www.synthfont.com/Downloads.html


P.S.
Unfortunately, the YEM (as stated by Yamaha) still does not fully support the sf2 soundfont format. So you need a bit of luck that an imported sf2 file actually sounds as expected. In some cases it can help if you load the sf2 file in an sf2 editor (e.g. Awave Studio, Polyphone or Viena) and just save it again.


Hope this helps!

Best regards,
Chris

[BogdanH]

The only benefit of loading soundfont directly into YEM is saving a lot of time -that way we can quickly get an impression of the sound (in sense, is it "interesting" or not). And even then, we should turn settings in YEM to default values -is the only way to "judge" the sample without being affected artificially.
The big downside of using soundfont directly, is huge resulting voice size. Most soundfonds contain sample for each note individually, which we don't really need. In YEM, we can remove samples for keys we don't need, but there's another problem: most soundfonts contain unnecessary long samples (which increases voice size tremendously) -and we can't shorten samples in YEM.
And the third potentially downside is sample rate. The thing is, soundfonts many time contain 48000Hz or higher samples (which again increases voice). If we manually add samples, then YEM only accept 44100Hz samples. However, if we add soundfont, then YEM also accepts higher sample rates.
That is, to get reasonable small voices (without actually reducing the quality), soundfont samples (for notes that we wish to have) must be extracted as wav files and then shortened and resampled. And then we import them into YEM manually.

It's interesting how differently we "judge" the quality of piano voices. I realize, that many prefer "nice sounding" piano voices... regardless of how real piano actually sounds. And I think that's the reason why piano voices in our keyboards sound as they do: they sound "nice", but far from real piano (in my opinion). Most built-in piano voices are over-polished, have too short decay/sustain, lack resonance, etc... and our ears adapted to that and accepted the sound as "piano sound". And when we find a soundfont that has samples sounding closer to real piano, we aren't capable to recognize that (unless we compare it to real piano).
About too short decay/sustain... for example: if we hit (ff) any key in C1 octave and keep it pressed, every real piano has very audible decay/sustain of over 30 seconds. Piano voices in our keyboards however, have max 15sec sustain, where the last third is at such low level that is not even worth mentioning. Some might say, nobody plays single note for that long time. Is true. But decay time affects how loud will a sound be after (for example) only 2, 3,.. 5 seconds -and that is important! This also applies for higher octaves notes, of course. How is that solved in our keyboard? By adding effects (reverb, hall, etc.)... which just isn't the same.

Just sharing my thoughts,
Bogdan

[Amwilburn]

Ah, thank you both, I'll look at the reverb settings, that might be it. BH, how do I check the sample rate and # of notes sampled? You're right, you absolutely don't need every note sampled or a rate above 44.1 (since the PSRs themselves don't use sample rates above that!). One of the Steinways was every 4th note sampled, I'll have to see if I can get rid of the white noise wash,

Mark

[BogdanH]

What you need, is Polyphone soundfont editor -it's free, full featured and easy to use. Additionally you need an audio editor, obviously. Many use Audacity, I prefer Goldwave.. any will do the job.

When you open soundfont (sf2) file in Polyphone, you will see four buttons on the left side. One of them is Samples. Click on "up-arrow" next to it and list of all samples will appear. If you select particular sample, then you can see all info about that sample (incl. sample rate). Here you can change sample rate if you wish (or later in audio editor). You can also change sample rate of all sample inside sf2 at once (click 1st sample, shift+click last sample). Once samples are selected, just pick sample rate value you wish to have (keep in mind, it can take a little while if you select many samples).
You can export samples in similar manner: select samples and click on Toolbox icon on top. You'll be asked in which folder you wish to export samples and that's it: wav samples are ready to be modified.

Now, sample names don't necessary stand for correct key.. for example, sample named C2 will actually be C0 key. To solve this mystery, click on Instruments "up-arrow" on the left side and select one item there. You will see a spreadsheet on the right side. Here you pay attention on top line (sample names) and on values in "Root key" line. Even there are a lot of samples, you'll soon recognize the pattern (because sample names usually are sorted properly). Btw. for one sample note, L and R channel are shown separately, but exported sample will be single (stereo) wav file.
After you export samples as wav files, I recommend you to rename samples according to root key -that way it will be easier to import samples into YEM later.
That was easier part so far and can be done relative fast.

Now you need to open each sample in audio editor. What you do there depends on situation, but usually you will need to shorten them (and resample, if you didn't do that in Polyphone). Can take a LOT of time... The rest is voice creation in YEM.

Hmmm.. I'm not sure if I answered the question...   
Bogdan

[Amwilburn]

Hmmm.. I'm not sure if I answered the question...   
Bogdan

yes, I think so! Thanks!

Mark

[lmederos]

I apologize for coming late to this party...

I would be very interested in the download link for these too, if  possible!

Thanks in advance.

[BogdanH]

hi Imederos,
Download links for free soundfonts are already given in this thread. Seems two most interesting are:
Lyzen's soundfonts and
Soundfonts 4U

Bogdan