Piano World Home Page Forums Our Most Popular Forums Digital Pianos - Electronic Pianos - Synths & Keyboards My free piano libraries in .sfz format

Hi,

some of you guys might be interested in checking out my free grand piano libraries at

https://luciphercode.wixsite.com/sofia-mz

There are a circa 1930 Steinway D, a 2020 Fazioli F308, a 1962 Bösendorfer Imperial, and a 1956 Blüthner Model 1.
They were all recorded at various location in Italy at 24 bit, 48kHz in 20 velocityy layers.

Hope you'll enjoy them.

That sounds really interesting! Thank you for sharing these instruments, and for bringing it up here.

And welcome to PW!

Thanks. My pleasure.

First, thanks so very much for sharing!

This looks super impressive and just from the specs alone it seems you’ve put a great deal of work into this. I’m always excited about new instruments, especially FREE ONES, and this one looks to be very promising.

Well done on the presentation, professionalism, website, and details.

One question: Does every key have 20 velocity layers or just some, since I see they are only 600 samples in total, per instrument. Or does the sfz transpose notes automatically since I see you’ve sampled in chromatic thirds? Just curious about that.


Also, if I may add a few gentle and positively-given suggestions:

1. It might help to have a few more demos of slower pieces. With super fast pieces, everything sounds fast and flashy, and shows the instrument can handle fast notes, but it can be hard to hear the unique tone and timbre of the instrument and those other subtle sounds, because everything is awash in the brilliant shower of notes.

2. The notes in the Steinway demo are wrong in a few places. Does this mean that the Steinway instrument may have some issues with note accuracy, or just the demo? This might just be an issue with the demo, but if it can be fixed, it might be helpful so others won’t think there’s an issue with the instrument.

Anyway, I can’t try these now as I’m away from home, but I will definitely give them a try and am excited to see what you have. And thank you for putting this together and sharing it with us!

Hi Taushi,

Yes, every note comes in 20 velocity layers, mapped to privilege low and mid range. I sampled only four notes per octave. Sampling 88 notes was beyond the scope of this project, but minor thirds usually work just fine for most pieces.

1) I agree, the demo pieces weren't the best choice. Someone was kind enough to record a slower and expressive track to test my libraries, which can be found at the following link:

https://linuxmusicians.com/viewtopic.php?p=161322#p161322

2) Thanks for pointing that out. I hope not LOL. But I need to do a better job with the demos, when I'll have time.

The sampling seems done in steps of minor thirds, so the sfz-player (eg. Sforzando) would transpose one half-note up/down to cover the whole keyboard.

But that leaves another question: was the sampling done with raised dampers, or with lowered dampers (i.e. pedal down, or pedal up)?

Thanks for the confirmation on that. How interesting that the sforzando player transposes it, on the fly.

My brain is still amazed by how quickly this stuff all works while our fingers are flittering over the keys, lol.

I’m also curious about how the sampling was done.

I’m going to give this a try later today when I get in and see how I like it. I must say, although the demos don’t really give the best idea about tone and timbre, the noise floor seems really low. I’m not picking up some of that background hiss like I get from other sampled instruments.

Will check it all out today.

If the same wrong notes appear in the demo files of the different pianos, that points to an error in the MIDI file that was used to create the demo. If the wrong notes appear in only one, that would point to an error in the programming of the sample set.

The way these systems typically work is that they stretch in the down direction (which sounds better). So the stretch is probably not a half-step up and a half-step down, but rather a half-step down and a whole step down.

This is generally how sample playback systems have worked. The idea of sampling every note for musical instrument reproduction is a relatively recent development. For a long time, memory was simply too expensive to make per-note sampling viable.

Incredible and outstanding effort! I just listened to the mp3s, and they sound fantastic. I haven't had the chance to thoroughly explore all of your sample sets, but I must say, the quality of your work is truly remarkable!


You've made some really clever choices with your hardware and software, and your sample editing shows great taste (or ears I must say!!!). I'm genuinely curious, how did you achieve this? If you don't mind sharing, could you tell me a bit about your background and any prior experience in this area?

Your work is incredibly professional, almost like it's on par with high-quality VST plugins!

I downloaded your pianos and have been playing around with them for a bit.

I have to say that I am very impressed - the sound is extremely clear and present and the playability is also quite good. I particularly like the Bluthner - lovely instrument.

In fact, I can think of several commercially available pianos I've tried that don't sound as good as these. Well done and thanks for sharing!

I’ve tried dozens of amateur pianos and I’ve never liked ANY of them … until these. Nice work!

I haven’t yet put them through the paces yet. I need to see whether they support half-pedal and repeal.

But the raw sound is spot on. S’good.

Indeed great work.

So, I was finally able to give this a try for about an hour or two…and…WOW. This is…easily one of the best VSTs I’ve played.

RECORDING: Recording is done very well, the noise floor is very low, and there’s not a lot of build up of hiss or noise from the samples playing concurrently. It’s recorded very clearly and well, and some EQ adds an even better dimension of clarity. Playability is pretty good, even though there is no velocity curve setting. The dry sound is excellent, because you can really hear the tone and timbre of the instruments, and even the “personality” and character if you will, since some of these are vintage instruments. Very well done for an independent person. These are better than some of the most recent and modern VSTs that were actually put out by companies dedicated to it.

DYNAMIC RANGE: The dynamic range is pretty impressive too. You’ve got some really powerful and dramatic fortissimos, and some nice piano to mezzo piano ranges. Pianissimo is ok, and would probably be easier to access with a velocity curve option. But the dynamic range here is superb!

FAZIOLI - The Fazioli was extremely well sampled and sounds great for a modern concert grand. This is what VSL’s Fazioli should’ve sounded like. I’m aware that Fazioli’s semi-concert and smaller concert grands aren’t always as wide tonally and dynamically as some other instruments, but an F308 is the crown jewel of their selection. It has a nice dynamic range and tonal variation, and that is actually captured here. This is what a Fazioli should sound like! I actually enjoyed playing this one the most, as the stereo image seems wider than the others, and the dynamic range is more easily controllable. This is apparently a 2020 Fazioli, so it’s still fairly new and extremely capable as a concert instrument.

STEINWAY - This was my other favorite. It definitely sounds like a Steinway, but it is clear this is an older instrument. According to the website, it’s an older vintage 1930 Hamburg. That said, it’s a very well maintained Steinway, and probably was rebuilt and refurbished over the years. The Steinway tone is definitely there, and the tonal palette is wide as you’d expect from a Steinway. The fortissimos, especially in the bass range and be crashingly forceful, so this instrument hasn’t lost its power. The pianissimos leave a bit to be desired, but…it’s still very, very nice. And again, I think a velocity controller might make it easier to explore the lower ranges of the instrument. This is a very faithful capturing of a Steinway, and I quite enjoyed it. ***There is a very minor issue with the sampling, see below for that.

BÖSENDORFER - This is definitely a Bösendorfer Imperial…for all that that implies. Love ‘em or hate ‘em, Bösendorfers have their own unique sound. That’s captured here. Luckily, it’s not as twangy and honky-tonk as some Imperials can sound, perhaps due to it’s age, since this is a vintage 1962 instrument. Dynamic range and tonal range is limited as is the case with Bösendorfers, but it’s not terrible and this is very playable, even to those of us who aren’t fans of the Bösendorfer lutelike, twangy, nasally sound. This compares favorably with, say, VSL’s Vienna Imperial. The bass isn’t as violently powerful and overtone rich as VSL’s offerings, though, which, again, may be due to the age of this particular Bösendorfer since VSL was using newer instruments.

BLÜTHNER - Another vintage model here from 1962, this Blüthner has the sound you associate with that brand. There is a delicate pluckiness to the upper octaves, a mute percussiveness to the treble range, and the lower octaves are more restrained than other brands as well. It’s faithful to the sound of that brand, and I think it’s a nice alternative to the more well-known Steinway/Yamaha/Bosendorfer tones. This had an ok dynamic range, comparable to the Bosendorfer, and the bass is even more reserved than the Bosendorfer, but still capable.

NOTES:

1. ***STEINWAY SAMPLING ISSUE: My suspicion that there were issues with the Steinway is correct. Two of the notes on the Steinway ARE incorrectly assigned, but only on certain velocities. The topmost F sounds as an E when played at certain velocities, and the topmost E sounds as Eb when played at certain velocities, hence why the demo came out the way it did. It tends to happen when played in the middle velocity range. I’m not an expert at this, so I couldn’t really identify which velocities, but it noticeably happened with those two notes. There may be other notes too. I would imagine this is a quick fix, though. Just a matter of making sure the WAV files are properly assigned and/or correcting the WAV files if need be.

2. If possible, add options to control the note-off and pedal samples. Sometimes I felt they were a bit too loud: as loud as the notes sometimes. I couldn’t tell which was the issue: the pedal samples or the note-off samples, but something there sounds like dampers hitting a note, but not always the actual note you’re pressing. I’m sure this adds to the realism, but it’s a bit TOO present. Allowing us to control the volume of those samples, individually is possible, might help us to identify where it’s coming from, and perhaps I could adds some suggestions. (To be fair, I EQ the gain on high ranges and treble ranges, because I always find that sampling of pianos is, by default, treble centric and flat simply because of how microphones work. So that could be why I’m picking up those note-off and pedal samples so profoundly. I’ll try again later.

3. Adding a velocity curve editor would be helpful to. (Is there one already in Sforzando player? If so, somebody please correct me). Leggiero playing (like Un Sospiro, for instance) is going to be difficult for come whose controllers have more weight, and one of the ways to circumvent this is to edit the velocity curve appropriately. I’d like to access the pianissimo ranges a bit more easily, also just for general playing. Is this something that can be done? (Now I didn’t use LiquidSFZ, which I see on your website is what these files are optimized for. Maybe that has one. If it does, please correct me. LiquidSFZ is an open source player and I kinda felt out of my depth with that one. I need a bit more research on how to get that to work. )

All told, though, this is EXTREMELY impressive, and a great service to the VST community. Especially to have done all this for free. Very nice!

To be quite honest, if you find a good piano, and comprehensively sample it: maybe 60 layers per key or 60 layers in chromatic thirds like you’ve done, and include comprensive una corda sampling as well…you could quite easily be an extremely well-done competition to even the big name VST samplers. I mean…this is very well done and impressive. Kudos to you and thanks.

Hello,

@Taushi,

Unfortunately, neither Sforzando nor the .sfz scripting language offer easy ways to implement velocity curve setting/editing.

If one would like to spend the time and tedious (!) effort, the piano's script can be edited to change which layers trigger at what velocities.

Other than that, .sfz is fun and offers a great option to build and share "homebuilt" VSTis like these, or the ones by Dore Mark, or the good-old Salamander Piano (Yamaha C5 in minor thirds/16 layers if I remember correctly).

A more convenient way would be to run Sforzando within a DAW or a host like Cantabile where you can create a curve that alters the MIDI input before it reaches the VSTi.

Cheers and happy homebuilts,

HZ

I can also recommend the velpro software which works very well , which allows you to have customised velocity curves per note and per region . Very simple and powerful and just appears to any VST as a midi input . Velocity per region is quite a good feature , depending how your DP is graded . Individual velocity curve per notes works well too and is useful when your DP sends inaccurate midi output ( sometimes the case with some Kawai and Yamaha keyboards ) it works better than the offset note adjustment in a VPC1 keyboard , as adding constant offset to velocity input to a particular note removes the possibility to trigger velocity 1 which allows silent sounds in rare situations when you need to press a note silent to trigger harmonics . I think Roland is the best in that particular area of midi output accuracy covering the whole spectrum, even on a FP10 ( with light mode selected)

Picking up from those other posts... really, velocity curves never exist in sample sets. Some VST pianos are more than sample sets... they are also their own applications with things like a graphic interface, velocity curves, ways to choose different subsets of the sample set (like mic positions), user adjustability of different sample sub-sets (your key-off and pedal noise concern), etc. But a sample set by itself doesn't handle any of those things, it all has to be programmed outside the sample set itself. I"m not a soundfont expert, but I think they don't include anything but the samples and their assignments, none of those other things. So then velocity curves would be handled by your host, though they may also be handled by an additional app or by your keyboard or by a device between your keyboard and your computer.

As for the volume of note-off, that can also depend on whether your controller sends release velocity information, or how its release velocity is set.

The more I play these, the more I like them - really exceptional. I'm particularly impressed by the fact that these were each recorded on site in different locations and yet you've managed to achieve a remarkable consistency in clarity and ambiance, while still capturing the individual character of each instrument.

I agree that finding a way to adjust the velocity response to fit individual keyboards would really be beneficial. The overall dynamic range seems decent, but I find it rather difficult to access the softest and loudest velocities on my controller - the majority of the response seems squished somewhere in the middle range.

I may have to check out the velpro software or some other kind of intermediary solution.

But again, just incredibly impressive work.

First of all, this is not Sound Font, the format is SFZ, Sforzando.

Second, SFZ is quite capable and almost does majority of the things you mentioned. SFZ requires a separate text file where you specify groups, velocity ranges etc...

You can even use its internal wave generator to build Saw, Sine, etc waves. I'm not sure if the FM module works, but I assume there is AN and FM, at least theoretically based on the documentation.

I'm sure about the Sine and Saw waves though!

"Before you cast the first stone, ensure you've mastered the subject inside out."

For adjusting the velocity curve in players like Plogue Sforzando (which doesn't seem to have a velocity curve editor), I just started using midiCurve; it seems to work pretty well. It gives a traditional graphical interface that you can make fairly detailed adjustments with, although I'm not sure if it can do individual velocity curves per note.

https://www.kvraudio.com/forum/viewtopic.php?t=493441

Scroll down to where it says

midiCurve (64-bit) can you find in the archive "pizjuce_x64_20120111.zip" from:
https://code.google.com/archive/p/pizmidi/downloads

You'll want the 64-bit version. Unpack the .zip file you download and find "midiCurve.dll", and move it to your folder with VST plugins. Then load it in your DAW in the chain before your piano VST.

Okay, instead of saying I'm not a soundfont expert, maybe I should have said I'm not a soundfont or sfz expert. But sfz is also typically seen as a soundfont format.

The "groups, velocity ranges, etc." that you mention is the same in either format (well, depending on how you define "etc."). With SFZ it is in a separate text document, with SF2 it is part of the same file that has the samples.

But yes, I know that sfz is capable of more than sf2. AFAIK, though, it does not support any of the things I said it did not support. The SFZ text file allows you to do lots of things but it does not include a visual interface, or any way to provide a user control for velocity curve, a user control for release or pedal noise volume, or a user control for alternate mic positions, or even "hooks" for some external player to provide the graphic interface for these functions. So, not only does it not do, as you put it, the majority of things I mentioned... I don't think it does any of them (the four things I mentioned, which I just repeated). If these things can be done at all, they would have to be done external to and separate from the SFZ format/document itself (which also gets back to my mentioning different ways you could apply a velocity curve). But if I'm wrong, instead of just saying "it can do these things," maybe you can point me to some info to support that? Because I believe everything I said applies to both soundfont generically as well as sfz in particular. Here are some references:

Differences between soundfont formats:

https://www.polyphone-soundfonts.com/documentation/en/manual/annexes/the-different-soundfont-formats

And here is a doc explaining the kinds of things can be included in the .sfz text file. They are using the sfz tern to refer to just the text file, but IME, most people use the term to refer to the combination of that text file WITH a file of samples, and it is that combination that is equivalent to, essentially, an enhanced soundfont (the SFZ text file itself not being anything that yields a sound).

https://sfzformat.com/