Help us improve the sounds in MuseScore

• Feb 27, 2018 - 17:33

Hello fellow MuseScorers! This is S. Christian Collins, virtual instrument designer and creator of the GeneralUser GS SoundFont. I have been asked by MuseScore to make some major upgrades to the default SoundFont for the next version of MuseScore (and beyond).

My goals for the new SoundFont include:
1. Improve the quality of the instrument sounds, starting with those most frequently used in scoring.
2. Address known issues with the current SoundFont, including notes that are out of tune, and the extreme volume difference between PP and FF dynamics.
3. Only freely-licensed samples will be used (e.g., MIT, CC0, CC-BY)

The plan is to gradually replace the instruments from the current FluidR3Mono SoundFont with higher quality samples and better programming. Instruments that continue to use FluidR3 samples will be updated to scale better with velocity changes (both volume and tone).

Development versions of the SoundFont will be shared on this forum to get feedback from the community regarding any changes we might make. I would like to avoid any functional regressions with the new sounds.

Ways that you can help:
1. Testing: Test out the instruments as they are developed, and let me know of any problems you encounter.
2. Create / Suggest Instrument Samples: One of the problems with using only free samples is that the available options can vary widely in quality. I will be posting to this message board with requests for samples as they are needed. We already have one generous user (ericfontainejazz) who has offered to record his saxophones.

I cannot guarantee that we will be able to use every sample we are sent. For every user-submitted sample included in the new SoundFonts, those users will be credited in the documentation that will accompany the release of the new SoundFont. The new SoundFont will published under a CC-BY-SA license, but many of the contained instruments will be usable under a less restrictive license. This licensing has no bearing on scores or recordings made with MuseScore, but only applies to those who want to redistribute the samples in some way. We are encouraging users to submit their samples under the CC0 license so that we can make as much of this "culturally free" as possible.

I will be posting a new SoundFont for you to try very soon, so keep your eye on this thread. I am very excited to be able to work with MuseScore--a project that I have long been passionate about--and I look forward to working with all of you along the way!


Comments

I know I'm kind of late to this game with suggestions, but I did have one I have been meaning to place here. As we all know, musescore is a notation software, not a production workstation. Yet, we all love to pursue realism with better soundfonts etc. What I'm about to suggest seems really pointless, but hear me out. Before using any DAWs I like to write things out in musescore, and I often try to get the musescore playback as close to what's in my head as possible. For example, when writing for guitar (or as a simulated guitar) I use a regular guitar sound and then I add an instrument and, using the mixer, replace it's sound with a midi instrument called "Fret Noise" which sounds like fingers sliding on guitar strings. And I write notes for it in wherever real guitar shifts would occur. I know I'm not the only one who uses musescore like this, but if I somehow am, that would make the following suggestion completely pointless. Hahaha

So this would be very tedious, and maybe even somewhat unrealistic, but once all the samples are selected for instruments like Violin, Trumpet, flute and any other instruments that have a tonguing/slurring, bowing/slurring, and maybe even plucking/slurring variant (though the latter seems less practical), perhaps slurred variants of those samples could be provided as well. And I'm not talking about finding new samples with slurs for every instrument. I mean using the same exact samples, duplicating them, and cutting of the attack portion of the note. But rather than completely chopping it off, you would have to fade the sample in, but quickly. Similar to how producers chop, splice, and arrange samples with out getting clipping. I'm not sure if this would be too hard to realize, but if it is as simple as I hope it would be, (but probably won't be with so many samples varying by instrument, range, and source) one would be able to duplicate the relevant samples, determine the proper amount time to fade it in enough to [not completely] kill the attack so that it sounds like a slur, and apply that to all of the samples. Even if it can't be done, I'd love to see it attempted.

Just a thought.

In reply to by s.chriscollins

That's what i am looking for : per-note expression !
and, is it possible to add the spiccato channel for strings in the instruments.xml distributed file ?
are ther plans to add more channels by default also for timpani (tremolo (roll)) ? all are in VSCO2 CE samples.
Thank you for your work.

In reply to by brouits

@brouits
It is possible to modify the amount of channels for an instrument in that file, but personally I find it tedious. https://musescore.org/en/node/267949 This was my struggle with that but I found that you can't program another channel to have an extra sound from, say an extra soundfont listed after whatever GM soundfont you're using. I don't think spiccato is any of the 128 OG midi sounds so that might not work right now. But hopefully these new extension packs will allow us to change the xml files with ease when they come. I'm not sure about more channels by default. Maybe Marc Sabatella or lasconic would have those plans.

In reply to by fernandoamartin

Nice :) When I implement legato, I plan to use a modulator to adjust note attack & sample offset in response to a MIDI Control Change (e.g., CC21), so separate presets won't be necessary. Then we can have MuseScore send CC21=127 when under a slur (except for the first note) for a more natural legato, and CC21=0 when outside a slur for normal note attacks.

In reply to by s.chriscollins

That's a very nice idea! You say "I implement legato". Are you a developer? If Musescore could play legato automatically under every slur that would be wonderful. In default midi, attack controller is on bank 73. But changing the attack time right to 127 may have an unpleasant effect and fast passages may become hard to hear, depending on the attack of the original samples.
When I use Rosegarden I move attack controller from 64 (the default value) to 74 or 84, depending on the soundfont I use.
When you implement legato, can you, please, leave an option where we could set different attack values for different staves?

In reply to by fernandoamartin

Yes, I am a developer working on the SoundFont. A convincing legato will be more about the use of sample offset than note attack speed, so it might not be possible for the attack to be user adjustable without creating undesirable side effects. My goal would be to get the legato behavior good enough that users wouldn't need to bother with adjusting it manually.

In reply to by s.chriscollins

While I ask about this here, I thought I might as well pitch another request I have. Let's say you load 'strings' in as an instrument (with the grand staff and using soundfont patch strings fast). When I set the strings to pianissimo, I find that it feels (or sounds to me) more like the upper strings are playing at a mezzo forte and the celli and basses are playing the desired dynamic. May it be readjusted? I'm not sure if this sort of thing would be a difficult modification, or if you're planning on readjusting levels on everything later (although this is the only patch I've found be less to my personal preferences) but if you could consider looking at that when you have time, I'd be grateful. [Also I'm not sure if I have the latest version of this or not, so it may be something you've already done [Considering I'm using musescore general full from late december]

Now my question about the legato...

So, would you have different legato settings for different types of instruments (in a way that would make more convincing legatos for most soundfonts of a given instrument (or family), or maybe just for this soundfont) or is it planned to be one size fits all legato for the whole soundfont player?

Thanks! really appreciating the work you've been putting into this

In reply to by s.chriscollins

I don't understand fully how it's going to work but I hope you succeed. When you implement this, please, let us know. (Currently I have been writing for keyboard instruments only because no matter how realistic soundfonts are I can't reach a convincing legato.) That feature will be very helpful.

In reply to by s.chriscollins

Hi @s.chriscollins!!!
I often use your soundfont file with MuseScore, because I get a lot of very close natural sound from it.
But... To me, there is a very great problem with the Violin, Viola and Cello (even Double Bass, but, because the pitch, the problem is not to be an audible issue, most of the time).
The problem is when we need a lot of high speed repeated notes; let's say: 64th figures at C4 to fill a half note.
We cannot to hear well each individual sound.
The result sound, is a notorious dull and week sound. Not a pleasant sound experience.
Can you, please, to work around this? ???
Another problem is about the bass pith notes of the Clarinet: they sounds with some notorious delay, not suitable when we use it with another instrument with high speed attack time.
Well, I hope you have an upgrading of the file, soon.
Blessings and Greetings from Chile!!!

In reply to by jotape1960

Hi jotape1960, are you noting these issues in the MuseScore_General SoundFont or GeneralUser GS? (or both?) I've tried to improve the slow low notes on the clarinet in MuseScore_General as well as the cello, the latter by replacing it with the more sprightly cello from GeneralUser GS. I am really hoping to eventually get new samples for these instruments with a quick attack + proper bass clarinet samples. I don't know that there is much I can do to help the existing sample set, I'm afraid, without making it sound even more artificial.

Is there a way to hear or have play back scoops and gliss and fall offs. Also a better way to notate them such as like sqwigally line for a fall off like in jazz music. Another thing I would love to be fixed is A way for play back with sfz, and Fp and etc. where it should hit hard and get soft and then crescendo up. I have tried to do this with dynamic and crescendo markings but it doesn’t do this with the play back. And lastly just dynamics play back needs to be better there are drastic differences then subtle depending on certain instruments and almost no changes on others causing me to sometimes have to have multiple dynamics in 1-2 measures to get the desired playback sound for MP3 files. Then have to go back and change them to have it look professionally correct to print the music off. Also with play back some instruments can not be heard at all or will get drowned out from the mass of the rest of the ensemble despite turning volume up on the mixer. For example the mellophones in my marching band music can’t be heard that I have them turned all the way up on th mixer and have them on FFF dynamic wise and their part tends to be covered or hard to hear and in reality that’s a very audible instrument in marching band. I’ve even tried playing with the pan options and reverb and chord options and it still isn’t quite heard. But just a few of my observations I’m sure you will give everything you do your best. And good luck to you and thank you for doing this.

In reply to by devon.mcdonald.779

You can fudge the dynamics by using the inspector to change the velocity values for that symbol, so that you can have the correct symbol for print-off but with the velocity adjusted to give the best available balance for the .mp3 rendering. You can also introduce additional "invisible" dynamics as may be needed for the playback, but again using the inspector to mark them "not visible" (they will be grayed out on screen).

I agree, the lack of dynamic modulation on individual notes is a major shortcoming of MS's synthesis model. You can try to fake it by using multiple notes, but each note may then sound with a separate attack. It works on tympani rolls using tremolo, breaking the rolled note up into shorter notes. Then you can use two voices - one for playback with the notes set "not visible", the other for print-off and marked "do not play".

I've been seeing more about 3.0 lately so I was wondering if this soundfont is still under development (more like making sure haha). I haven't seen any news about it in a while, but I'm still looking forward to it, whenever it comes.

In reply to by speedmeteor101

The SoundFont development took a back seat for a bit while I created the MuseScore DrumLine (available through Help -> Resource Manager). I'm also in the process of moving to a new home, so after that, development will pick up again. Thank you all for your patience!

A new version of MuseScore_General will be coming out soon that will feature fixes to reported issues (e.g., out of tune choir) and updated marching percussion sounds backported from MuseScore DrumLine. This will be followed by some delicious new ensemble strings with individual Violins, Violas, etc.

In reply to by s.chriscollins

The marching snare has never worked. I have musescore installed on Windows and Linux and it works in neither case. I do hope this gets fixed.

Sounds aside, one of the worst problems is not particularly with the single basic soundfont, but the synthesiser. It has the capacity to add soundfonts alongside the pre-installed one (which is VERY necessary for particular tailoring of output, but the handling of these is frankly atrocious.
They all end up in one giant, unwieldy list which turns your brain to mush when trying to find things. Plus setting the order via the arrows is slower than watching paint dry and makes my i7 (yes really!) with 16GB of RAM unresponsive. That should not be happening.

Every time an update arrives it just wipes out all those soundfonts leaving a carefully constructed score playing a bank of 'grand pianos'. It's impossible. To counter this I have to run up scores on a laptop not connected to the internet with an untouched copy of Musescore and instruments.

The support for sfz files via zerberus is generally problematic. Having these loaded into zerberus drastically delays startup and on linux causes me to have to restart again before any scores will even play.
I posted this to musescore forum some time ago, but the two permanent, voluntary oracles there, whose knee-jerk response is to ask for files, don't realise that almost everyone personalises musescore and, importantly, by using the features intended for doing so; or they wouldn't be there for that purpose. I admit I gave up on that forum because of this issue.

I see it thus: musescore is designed primarily for use with one, single soundfont bank, which is not, and never has been, satisfactory for a good number of users. Many improvements have been made (some regressions). With that design the ability add further soundbanks appears as an afterthought. Some more ability to organise these soundbanks would be greatly appreciated.

In reply to by Roger v.d Velde

Marching snare should work fine in the default soundfont, it certainly does for me. Probably the issue is with some non-standard customization you have done, either via a template, a style file, a drumset definition, or your instruments.xml. In order to assist you better, please start a new thread and attach a sample score to demonstrate the problem you are having, and we can go from there.

In reply to by Roger v.d Velde

Absent a specific example, my guess is that you are being fooled by the normalization that is applied. That is, MuseScore is probably applying the soundfont just fine and using the pianissimo sample correctly, but if there is no louder passage in the score to set a peak level, then the exported mp3 is getting its peak from "pp". This is, well, normal, as it is consistent with how audio is ordinarily produced, but if you need a quieter mp3, you can always reduce the volume in an audio editor.

In reply to by Marc Sabatella

Your unwavering belief in the total infallibility of musescore's operation is admirable, but misguided. Using regular sf2 violins, which are more stable, but inferior-sounding, the dynamics can be exported perfectly fine. If you take the trouble to see what is said about the support for sfz in the musescore user-guide you'll see that it is only supported as an afterthought, not as a stable system. The dynamics applied in the score do not export. Unless you've actually tried this you cannot know and can't have a useful opinion regarding it.

Having it as an option under those circumstances is frustrating. It's like being told you can bring your own sandwiches to a concert, only to find you have to eat them in the toilets. You may feel that I am rather brusque when addressing you, but the only issue is that you seem to think that attaching sample scores is somehow the solution to every issue. It may be for some, but issues like limited support for sfz, it most certainly isn't. Someone at musescore central - and you yourself if you care to - could try it out and figure out the problem. That, after all, is what the software developers there are meant to be doing right? Or is partially-functioning software okay now?

In reply to by Roger v.d Velde

Actually, I've fixed enough bugs in MuseScore to know very well it is anything but infallible. But I've also helped enough users here in the forum to know that often people are just confused about how things work. And the particular issue of dynamic / audio normalization is one that does sometimes catch users by surprise.

So, as I said, absent a specific score and steps to reproduce a perceived problem, it's hard to know if there is a bug or just a point of confusion. So we are limited to just guessing based on prior experience. And since export of dynamics work fine for me whether using SF2 or SFZ, and I've heard enough examples from others to know it seems to work fine in general, and this normalization thing is a pretty common misunderstanding, I made my best guess.

Now, I'd much rather fix problems and/or truly help people than guess. Which is why I am so often encouraging people to provide the sort of additional information we'd need in order to do either.

So, again, if you believe there is a problem here, I'd love to help. Which is why I continue to urge you to post to an appropriate thread (eg, the one you started some time ago) with the requested info - a sample score and precise steps to reproduce the problem. Because I have tried it, and it does work fine in general. I even tried it again just now, and it worked as it always has.

BTW, while it is true than in older releases SFZ support was limited (to just what was needed for the Salamander piano), this has improved gradually since then to the point where we now can do considerably more with SFZ than we can with SF2.

In reply to by Marc Sabatella

The 'customer service' tone is quite annoying. the problem isn't 'perceived' or one I 'believe' I have, it is a reality. There is no 'confusion'; not from me at least. Please stop asking for a sample score, you don't need it, it will do you no good.

I will however post to another thread and conduct my own survey concerning dynamics applied to sfz instruments specifically - and the ones in question - and their output to mp3. I would like to see if, and how many, other people actually have a similar issue. Perhaps these people will also be able to offer some useful insight into it rather than asking for sample scores.

In reply to by Roger v.d Velde

If the developers can't reproduce the problem, how do you expect them to fix it? If they had a copy of the problematic score, perhaps they might be able to trigger the bug and find out what is going wrong. Also, if you haven't already, you should report this bug on the issue tracker: https://musescore.org/en/project/issues/musescore

The developers will want to know info about your system (OS, MuseScore version), etc. to hopefully help narrow down the problem.

In reply to by s.chriscollins

They can easily reproduce the problem by running a clean copy of musescore like I did, with only the solo strings from virtual orchestra added alongside the standard soundfont then applying dynamics and seeing if they output to mp3. Very, very simple. I've set it out several times in plain English.

I don't see what's difficult to understand about this.

In reply to by Roger v.d Velde

The difficulty, again, is that we have done this many times, and it works fine for us and for everyone else. In order to understand what is different in your case, we really do need your help, as we quite simply have been unable to reproduce any sort of problem. So thanks for starting the new thread where we can discuss this issue further.

BTW, note I didn't say you were confused, I said many users do find this normalization to be confusing. I'm glad you understand it, though!

In reply to by Roger v.d Velde

Roger, thought I would jump in here and help to clarify a few things.

Historically, the priority of the MuseScore development team has been to create an incredible solution for engraving. MuseScore does allow you to create beautiful scores for print and MuseScore 3.0 will do this even better and faster, with some great auto-layout capabilities.

Now that the software is becoming much more mature and capable, and now that there are many additional resources available, playback is becoming much more of a priority.

The current playback limitations are not only well known, but very clearly defined, and there is a plan to continue to expand this capability as we move forward.

To address the issue of SFZ playback, specifically, Zerberus currently supports a very limited number of opcodes in the SFZ spec. Many new ones were added since 2.1 and many more will continue to be added in the future. As additional opcodes are added, this changes the way MuseScore is able to essentially "drive" the specific soundfont.

As each unique soundfont may utilize different SFZ opcodes, a very wide range is needed to support more than the most basic aspects of SFZ (ex: to utilize the full range of expression/articulations for strings could require supporting a considerable number of opcodes).

In addition to this, performance is also greatly affected by the ability to support disk streaming vs. storing all samples in memory. This is also something planned to be addressed in the near future.

Specific SFZ opcodes currently supported in Zerberus vs. Linux Sampler or Aria Engine - https://docs.google.com/spreadsheets/d/1xZ8zAM5Hd-NljbpSFME0qH0aWSVXzFo…

The next wave of improvements in SFZ support will be for opcodes that will specifically expand capabilities for percussion, as this is a near-term priority. Many of these improvements may also be relevant to supporting other publicly or commercially available SFZs.

The most constructive way to get involved and the most productive way to encourage support for specific SFZs is to identify the specific opcodes required that are not currently supported and make a feature request to add specific support.

Here are examples of what those might look like:

https://musescore.org/en/node/273151
https://musescore.org/en/node/274258

In reply to by Daniel

Not to qualify the compliment I want to give to those responsible for MuseScore_Generalsf3, but I freely admit I am new to soundfonts, notation software,extensions,imports and all the rest which trifles with my sanity, but I have been in music for 5 decades and know sound. These fonts are much better than I expected. For my own work I have explored many sounds. For classical music brands like Vienna Symphonic seem to stand out. There are others which seem to be popular with film score fans, but these are lacking in bringing forth the relatively more subtle string techniques found in classical writing. All brands seem to have serious limitations (thus far). However, with my limited experience, I still feel yours stand out. Now,I have yet to have utilized MuseScore fonts for symphony orchestra having not scored here for more than 15 parts. So my observation is limited.... I'm sure others will differ.

I'll offer a few examples. I've just completed a song cylce which I've played back. The piano sounds like a piano. The dynamics and articulations work out. (The lows are penetrating.) The voice (mezzo) sufficed, though "she" was flat on high sustained notes, which rather makes the composer reach for the aspirin. (on the few notes just below middle C, she becomes a he.)
A ballet score is almost completed, one for 6 players. The sounds are fine with some suggestions I'll note. Once I played with the mixer, to distribute the instruments as they would be in the stage/pit setting, the playback was true.

The suggestions, which I'm sure you are aware of: Straight (and other) mutes for brass.
I've already mentioned the voice tuning.The marcato attacks of winds and brass could be sharper, especially the trombone. While the clarinet's chalameau is awesome, the upper register lacks some of the brilliance one would expect from a f or ff passage. The flute is perfectly malleable and the breathy lows surprised me. Of course, during slow, long notes the vibrato does wobble a bit, but I realize it is engineered sound.
The violin surprised me, too. The legatos are fine. As with all string attacks in fast passages, they are not as clear as we would like. I also noticed that during passages where the violin is solo, the sound is true, but when other instruments join in there is sometimes that "sameness" sound which I've noticed in MIDI instruments, as though it takes on the character of others. Of course it is difficult to temper the high notes (harmonics) to keep them at a pp.

I do not know how you reproduce these sounds, but I think your efforts pay off handsomely.
There is a particular joy for a composer to hear a work not yet performed played back with virtual fidelity. And I realize improvements are forthcoming.

And, like the MuseScore notation software...it's free.

Regards

The new MuseScore_General soundfont version 0.1.3 has now arrived in Debian unstable and the official PPA.

The soundfont is now split into two, the normal MuseScore_General (musescore-general-soundfont in Debian) which has become even cooler, even more featureful, with not just the new piano (from 0.1.1) but also new strings, and, more importantly, ensemble strings (to avoid the sample doubling effect)… unfortunately this means it’s now about four times the size the old Fluid (R3) Mono soundfont (fluidr3mono-gm-soundfont) was, so there’s now also a MuseScore_General_Lite (musescore-general-soundfont-small) which has comparable sizes (on-disc and, more importantly, in RAM and startup times) to Fluid (a bit smaller actually) while being of the same quality of the full-featured font, carrying the same presets, banks, etc. — the difference is that the light version does not actually ship distinct samples for the ensembles (so second violins sound exactly the same as first violins), and that it uses the piano and string samples from Fluid, not the new ones.

However, the light soundfont contains all the improvements (such as the Choir Aahs and some other instruments being slightly better in-tune) from the full soundfont, and the Fluid piano and strings also got improved. Furthermore, the MuseScore_General soundfont contains some stereo samples (yes, even the light version, which is impressive, as it’s smaller than Fluid (R3) Mono which was made mono to save space).

I’ve created the Debian packages (all three of them: Fluid and the two MuseScore_General ones) in a way that makes them installable on all kinds of Debian-based distributions, even positively ancient ones, so feel free to install them. (Do not, however, install packages from one distribution or version on another normally, and especially do not use the PPA on Debian. The soundfont packages are the only exception, because they basically only ship data.)

Non-Debian-Linux users can go to https://people.debian.org/~tg/sf/ and download the individual soundfonts for a sneak preview. Symlink or copy either of the three to MuseScore_General.sf3 (in /usr/share/sounds/sf3/ on Debian, /usr/share/mscore-2.3/sound/ on stock installations) to use it as default soundfont.

In reply to by mirabilos

Thank you!

This sounds good, but...

Can I install on my ver. 2.3.2? I saw a few unpleasant words in the description like "unstable" and in the PPA like "unsupported" and "untrusted"....and, of course, I have no idea what "debian" means.

But I would like a mezzo (choir aahs), etc, who stays in tune on high notes and a string ensemble, etc...

Also...would the new MuseScore general replace the old so the old should be deleted from the synth?

In reply to by penne vodka

OK, let me walk you through a few of the “scary words”:

“Debian” is a kind of Linux. It has a “staging area” called “unstable”, in which packages get uploaded and stay there for a couple of days, until they migrate to “testing”, which is what will eventually be the next “stable”. This is mostly a process depending on time, and new stuff always goes to “unstable” even if it is stable, it will just migrate smoothly; this gives users the chance to test and report bugs. (This has no bearing on the soundfont, more on real software).

So, yes, you can download and use this on your version 2.3.2 — if you want the full experience, just go to https://people.debian.org/~tg/sf/ and download the MuseScore_General_Full.sf3 file.

Then copy that into your MuseScore sounds directory as MuseScore_General.sf3 replacing the old one. This way, you will not have to change anything in the synth; it’s a drop-in replacement.

In reply to by mirabilos

By suggestion of S. Christian Collins, the SF2 of the full soundfont is now also included. It’s huge, but it completely eliminates the long startup times and possible audio compression artefacts. This is now also in Debian (as musescore-general-soundfont-lossless); I’ll update the PPA later.

The musescore-general-soundfont-small package also finally entered Debian today, so we can have these in backports in another fortnight or so (plus backports-NEW queue waiting time).

In reply to by penne vodka

I think it's always possible to use the former soundfont from here: https://musescore.org/en/node/278624#list (or should be possible).
I don't think the updated soundfont is available with the actual release or actual nightly at present (I didn't check it, but I don't see a difference concerning the size of the file) - maybe in future. And I can confirm, that I don't notice a long startup with the actual soundfont so far.

In reply to by kuwitt

@penne vodka: the SF2 is the same as the SF3, except for three things:

  1. It takes up about six times as much disc space (almost ½ GiB)
  2. It makes MuseScore start up instantly, instead of having to wait for sample decompression
  3. It avoids compression artefacts that some audiophiles may hear

So, for the average user, they want the SF2 if the SF3 makes MuseScore too slow to start for them, or the MuseScore_General_Lite.sf3 (1/20th the size of the SF2!) if they have small hard discs, but the Lite one lacks the new piano and strings.

@kuwitt: This is version 0.1.3 of MS_General; the page you link has only version 0.1.1 which lacks the new strings, but also many bugfixes, such as slightly less out of tune instruments and choir, and other improvements.

Yes, it’s true that the MuseScore 3.0 release ships with the older soundfont. That’s why I’m uploading the latest soundfont at https://people.debian.org/~tg/sf/ in the first place, so users can get it if they want.

In reply to by mirabilos

I'm using the 2.3.2 rev 4592407 AppImage in Linux Mint. When I view the Fluid tab in the synthesizer, it tells me MuseScore_General.sf3 is being used. My question is: is this the 0.1.1 0r the 0.1.3? Is there a way to tell? I can't see inside the AppImage file to get the size of the .sf3 file. Or does it even matter?

You say that "the MuseScore 3.0 release ships with the older soundfont" - does "older" refer to 0.1.1 as opposed to 0.1.3?

In reply to by mirabilos

I’ve updated the site https://people.debian.org/~tg/sf/ with the latest soundfonts:

  • FluidR3Mono_GM.sf3 is (still, unchanged) fluidr3mono-gm-soundfont version 2.315-5, SF3 compressed (lossily)
  • MuseScore_General_Full.sf2 is now musescore-general-soundfont version 0.1.7-1, upstream name MuseScore_General_HQ, SF2 uncompressed (lossless)
  • MuseScore_General_Full.sf3 is now musescore-general-soundfont version 0.1.7-1, upstream name MuseScore_General_HQ, SF3 compressed (lossily)
  • MuseScore_General_Lite.sf3 is now musescore-general-soundfont-small version 0.1.6-1, upstream name MuseScore_General, SF3 compressed (lossily)

The *.copyright files can easily be mapped by the Debian package names shown above. The individual soundfonts also contain full version and copyright information thanks to a small script I wrote that patches the RIFF INFO chunks at package build time. (Do remember that, while, as it comes under the MIT licence, it can be used in most settings, waveforms generated using this soundfont are “copies or substantial portions of the” soundfont.)

Important: MuseScore_General_Full now maps to MuseScore_General_HQ, the old “MS_General”. This is now an optional download for MuseScore otherwise. MuseScore 3.1 will ship with the newly-renamed MuseScore_General, formerly MuseScore_General_Lite, by default. (However, due to package naming and other constraints, Debian users will, by default, get the SF3 version of the HQ soundfont, unless they already have a suitable soundfont installed. I could only change this by shuffling things around I’d rather not touch, plus we’re in deep freeze and about to release the next Debian stable 10 (“buster”) any moment, so I refuse to follow this latest renaming.)

It should be noted that Debian 10 will not contain these new versions of the soundfont (but I’ll most likely provide them via backports), as it’s simply too late to bring in a new upstream version (Debian stable release policy). I am currently updating the PPAs, though.

Hello,
I'm a jazz drummer and noticed that I can't create any scores which use sweeping brushes - there is just no percussion font that includes that. It is very common instrument for slow ballads and tunes. I'm very surprised no one notices this, I'm guessing it's because vast majority of MuseScore users are classical musicians.

I did create sound brushes samples using my drum kit and incorporated it into standard .sf2 font
supplied with MuseScore. I have ability to record samples and willing to provide them to your team for free to all's benefit (I think it's CC0 license).

If MuseScore team is interested getting this flavor of percussion into the soundfont, let me know what do I need to do. At this point I just have sound samples. If there are special requirements to record them, I'd like to read more on this (provided, currently lacking sustained brushes sound is something desired/needed for updated MuseScore fonts, else if there is no plan to consider this, it's just waste of time recording them).

Please let me know!

Thanks,
Victor

In reply to by VT

Hi Victor. Thanks for offering to help out with this! As you may have noticed, the FluidR3 brush samples in MuseScore_General.sf3 sound very little like brushes. The sample for "Brush Kit" MIDI note 40 (brush swirl) sounds more like a hit than a swirl, and MIDI note 38 (brush tap) and 39 (brush slap) sound more like stick hits than brushes.

I would be interested to hear the samples you have created. The CC0 license would be perfect. Ideally, we would like to have the following samples at soft, medium and loud dynamic levels to replace the current brush kit, but we might be able to make good use of what you've already recorded:

  • Snare Brush swirl (single) - a single swirl of the brush. If you can get multiple samples of this at each dynamic level, that will allow us to potentially use round-robin to keep the sound from being too static. Optionally, also record a sustained swirl pattern at a slow tempo, and I might be able to chop it up for round-robin sample triggering.
  • Snare Brush stir (sustained) - I can create a loop. Try to avoid rhythmically pulsing the swirl so that it will work at all tempi.
  • Snare Brush tap - produced by making a quick down-up stroke, playing off the drum head.
  • Snare Brush slap - produced by playing the brush flat, striking down into the drum head
  • Hi, mid and low toms - brush hits on all tom toms
  • Cymbals - if possible, brush hits on crash, ride and hi-hat cymbals along with a sample of closing the hi-hat with the foot pedal

The most important of these would be the snare samples, but it would be nice to have toms and cymbals as well to complete the set.

Another thing to consider in all of this is the way that brush "stirring" is typically notated. If I recall, often text such as "stir continuously with left hand" or simply "stir" will appear over the drum staff. I'm guessing that in MuseScore, one would need to enter and then hide notes to trigger the stir/swirl samples, as there is no system in place that can create an automatic stirring effect with rhythmic, BPM-synced circular motion. Being that you are a jazz drummer, perhaps you could shed some more light on brush notation conventions, as my own drumming experience doesn't feature much jazz.

In reply to by s.chriscollins

Hi Chris, sorry for delay with this. I've recorded a few samples of brush sounds. How do I submit the file to you? It's 5.9 MB file in .mp3 format, OK for an email attachment or I can upload it to any specified location. I'll provide more details about the sounds in a readme file sent together with it.

So please let me know what do you want me to do next, how to proceed at this point.

Victor

In reply to by VT

Excellent! Do you have the audio in WAV or FLAC format rather than MP3? I'd like to avoid the lossy compression if possible. E-mail attachment should be fine. If the WAV/FLAC is too large for e-mail attachment, do you have a Google Drive or Dropbox account to upload it to?

In reply to by s.chriscollins

I got original .wav file, it's recorded at 48kHz bitrate 24 bit resolution by Zoom Q2n recorder. I don't have a dropbox ir google drive account, I suppose I can create one, but then I have to send you credentials to grab the file, do I? Sorry, never had need to use these tools, so don't know much about how this works in detail.

In reply to by Louis Cloete

Also, both Dropbox and Google drive offer the ability to share the file with another person. You won't need to give your credentials for me to access the files, just a sharing link/e-mail. Both services offer enough storage space for free for you to be able to share large files, and you can always delete the files from the service after the person you are sending them to has downloaded them, which will free up space for other files in the future.

In reply to by shoogle

I found the original poster of the soundfont, he sampled his own instrument and posted it for free in his site and on a facebook group but it has no particular licence i think, if you write to him he will surely allow its use for musescore. The soundfont is great, its better than many paid vsts out there.
https://selecciondefinitiva.com/2015/12/07/bandoneon-en-formato-soundfo… (this is where it was originally posted)
https://www.facebook.com/adriancharras (this is the posters personal facebook page)

In reply to by [DELETED] 26799858

People are very busy these days, so it's worth sending another message just in case he missed the first one, or saw it and forgot to reply. Just say something along the lines of "Hi, I was just wondering if you saw my first message about licensing your soundfont for use in open source programs like MuseScore? Doing so takes minimal effort on your part and it would make a lot of people very happy, but it's totally up to you and I understand if you don't want to.".

In reply to by [DELETED] 26799858

Thanks.

I fear you explained it wrong. CC0 does also allow usos comerciales, and it means the author gives up the right to collect money for it, it does not prohibit others from doing so.

Judging from the answer, though, Adrian seems to only be interested in people just using it without annoying him, so… it’s a very weak permit.

In reply to by mirabilos

I don't think this is correct. Under CC0, the author still could make money off of it if he wanted to, but so could anybody else. It's essentially in the public domain, and anybody can use it pretty much however they please, except to restrict others from doing so.

In reply to by [DELETED] 26799858

I see, i dont see how anyone would spend money on something on public domain they could get for free... Except they want to donate or something, i didnt say commercial use is prohibited. I can write back to him but as you can see its unlikely i even get an "ok" as answer. If this permit doesnt do it i think thats it. Maybe you can list it in the soundfonts page as alternative third party soundfont.

In reply to by [DELETED] 26799858

A better idea is to pick a license for him (e.g. CC0) and say:

Sorry but we need you to explicitly place it under one license or another. What about CC0? This will allow MuseScore (or indeed anyone) to ship it for free. Full terms are at https://creativecommons.org/publicdomain/zero/1.0/. If you are happy with this, all you need to do is type "I release this soundfont under CC0" and then we can use it. Thanks!

In reply to by mirabilos

A concise way to explain CC0 in simple terms is that it releases the creative work to the world to use without restriction. Many SoundFont designers essentially are intending to do this with the samples they share--they simply don't care what people do with them. This will often be evident in the language they use when sharing it, or they will say it is "public domain". Having a sound designer agree to use CC0 makes these terms unambiguous, and we can avoid any misunderstandings down the road. In addition, we credit all known SoundFont contributors in the documentation accompanying MuseScore_General, even though this is not a requirement of the CC0 license.

Here is version 0.1.4 of MuseScore_General. Please note that I have changed the bank assignments for the new ensemble strings. This affects all ensemble Violas, Celli, and Basses presets. Any scores using these strings with version 0.1.3 will need to re-assign the presets in the mixer upon switching to 0.1.4. I apologize for the inconvenience. This was done to better accommodate future expansion, such as the upcoming "expressive" presets that support single-note cresc. and dim.

Features of this release:
* Fixes the following issues: #281732, #280907, #280904, and #272992.
* Improved the dynamic response of the acoustic pianos.
* Re-numbered ensemble strings bank numbers to make room for future expansion.

In reply to by s.chriscollins

Thanks very much! I had a quick listen and it sounds great!

Forgive my ignorance, but could you please confirm or refute the following statements:

  1. Your soundfont (MSG 1.4) preserves the distinction between singular (solo) and plural (ensemble) instruments (i.e. "viola" uses a different sound to "violas").

  2. You added new instruments, but all of the old instruments are still included, and are in the order prescribed by the General MIDI standard.

  3. If we stick to using the old instruments then our scores will render as expected with any soundfont, but we are limited to what is specified in General MIDI (i.e. we must use generic "Strings" instead of "Violas").

  4. If we "opt-in" to using the new instruments (such as the separate sounds for Violas, Celli and Basses) then we gain better sound with your soundfont, but we lose compatibility with other soundfonts. (I understand that this is inevitable because those instruments do not appear in the General MIDI standard).

Is there any agreed (or de facto) standard for instruments beyond those defined in General MIDI?

Edit: so those are pretty basic requirements and having checked a second time I see that they are indeed all true. I must have done something weird the first time around because it looked like some of them were not true, hence I felt it necessary to ask.

In reply to by shoogle

All are true except for #4. If you use any of the new ensembles (Violins Fast, Violas Slow, etc.), they will render according to GM bank fallback when using a different SoundFont.

Let me explain how this works. All of the "fast" string section presets, for example, are on different banks but using the same preset number, which is 48. Currently it looks like this (bank:preset):

000:048 Strings Fast
020:048 Violins Fast
025:048 Violins2 Fast
030:048 Violas Fast
...etc.

If MuseScore tries to play 020:048 (Violins Fast), but that preset is not in the SoundFont, it will instead play using 000:048 (Strings Fast in GM standard). This of course also applies to the ensemble instruments using presets 44 (Strings Tremolo), 45 (Strings Pizzicato) and 49 (Strings Slow).

It is in this way that the new instrument sounds will be backwards-compatible with other GM banks.

In reply to by s.chriscollins

The highest bank number is 48 in the Roland's Sound-Canvas bank list.
For this reason, it's better to start the banks from the 60th number (instead of 20th).
Otherwise, banks 20., ..., 30., ..., 40., ... in other sound-fonts, will clash with the sounds in the MS-General's banks.
It's also not possible to replace the already existing banks with new ones in other sound-fonts (Because we can only do a two-dimensional definitions: bank-number:preset-number.)
I think: Bank 60 and above is ideal.

Some examples:
020:48 Orhestra4 *
024:48 Velo Strings
032:48 OctStrings1
033:48 OctStrings2
034:48 Contrabass Sect.
040:48 60s Strings *

024:50 Tron Strings
025:50 Noiz Strings *

024:52 Chorus Lahs
032:52 Chorus Aahs
033:52 Male Aah+Str

020:53 to 40:53 Voice Oohs variations * * (eg: 20:53*, 40:53*)
00:56 to 32:56 Trumpet variations (eg: 025:56 Warm Trumpet *)

Including Brass, Pan Flute, Synth Lead, Saw Wave variations and more.

In reply to by s.chriscollins

@s.chriscollins

I'm curious of what samples you'll be using for the solo strings. For violin I actually like the idk1609 samples that were posted for free by the user... and were actually used in the Alpine violin set. (Although the attack isn't the greatest on notes around middle C). I haven't really explored other samples, but a lot of cellos sound decent in my opinion and there's always a good bass out there. (I don't really know how a good viola sounds hahaha)

As for other specific ones, I don't have any in mind. I know that with the current solo violin in 1.4 and with aaviolin for example, that there are some overtones/ frequencies in there that just irk me a bit as they sound either too electronic, or too wet, or... well I think you get it. I have FL studio, so I was experimenting with the parametric EQ 2 (to target the freqs) and Maximus (compression), and actually did see strong improvement with the samples tone-wise. However, I'm no sound engineer so I don't know exactly what to look for to get it just right.

Even though I was applying it to the whole soundfont through the soundfont player/ mixer, I was thinking the same could be done for the samples themselves to build even more on the soundfont. I'm not sure if you do equalization and compression or if you know anyone who does, but I think it could GREATLY help the tone quality of the samples just from my experimentation to add even more.. I guess I wouldn't say realism but... [insert word] to the soundfont.

You've done amazing work with everything so far, and honestly I don't know if you've been doing this already, but I just wanted to let you know what I found through this experimentation.

Sample rate

Which sample rate do the individual samples use? Is it the same for all of them, or do they all differ, and if so, do they have a common divisor (e.g. 22050 and 44100) or not (e.g. 44100 and 48000), and what’s the most common sample rate?

Or perhaps: would it make sense to add a “sample rate” column to the CSV table listing the sample sources? Assuming that, at the very least, within one instrument the sample rate does not differ.

I’m asking because I just noticed that MuseScore offers changing the output sample rate at two places, and that they have different defaults: playing on the speaker from the application defaults to 48 kHz (with ALSA), while MP3 export defaults to 44100 Hz… (and WAV, FLAC, OGG Vorbis export either uses what is configured for MP3 or is not configurable at all?)

I assume it would be best to use the same sample rate for the output that my input samples have (with the table, I could see what the instruments my score uses have).

In reply to by mirabilos

There are different sample rates used within the SoundFont, most commonly 22050, 32000 and 44100, but others exist as well, including a handful at 48000. You won't gain anything by attempting to match output rate to sample rate, though, since samples are frequently not played at their recorded pitch anyway. Most instruments use a single sample to span more than one note, and only one of the pitches in that span will actually play at 44100 Hz (assuming it isn't also fine-tuned within the instrument). Fluidsynth's high quality interpolation avoids any aliasing effects from these sample rate changes. As long as your output sample rate is sufficiently high (44100 Hz or higher), it should all sound good.

I've started working on some SFZs for use in Musescore. I'm testing the new single note dynamics feature in the nightly build. However it seems that zerberus doesn't support the xfin_loccN and xfout_loccN opcodes so how does one create an SFZ that can make use of the single note dynamics?

Please can you load a sample soundfont that shows us how to make the definitions for CC02 and CC11 in a soundfont.
for example: Only with sine wave sound, and only with modulator parameters.
Because someone who wants to make a soundfont compatible with MS3.x needs this.
Because the current soundfonts (General and HQ) are very complex, I think that such a simple definition is necessary.
(The tests also give different results. I'm confused)

In reply to by Ziya Mete Demircan

Here is the simple way to make your own SoundFonts work with MuseScore's expressive mode:

  1. Edit your SoundFont in version 2.1.1 or later of Polyphone.
  2. Within each instrument (not preset), add the following modulators:
    modulators.png

The first modulator disables the default velocity-to-attenuation curve (notice the text saying "disabling default mod."). The second modulator recreates that curve with MIDI controller #2 as the source.

I used a more complicated variation of this technique in MuseScore General that uses a different attenuation scale and allows CC2 to affect the filter as well. In the strings, I even use CC2 to crossfade between the forte and piano samples for a more natural change in tone across the dynamic range.

In reply to by bottrop

Thank you, bottrop! I know what you mean, but the PC is my stereo.

What I was referring to was a review of desktop speakers w sub-woofer which were designed for classical music, but I could no longer find the article on the net. So...I got stuck with Logitech speakers which might be fine for games and rock, but not for what I do. After fussing with mixer and synth I get a good sound on my headphones, which are good, but now and then I want to be free from the confinement of headphones.

I could spend the bucks and get Bose or Klipsch, I suppose.

In reply to by jotape1960

You can make "normal" SoundFonts use single-note expression by setting MuseScore to use CC11 only. However, this will affect all loaded SoundFonts--you can't select different methods per-SoundFont, and will reduce the quality of the dynamic expression of the default SoundFont (MuseScore_General).

Please do not ignore the different possibilities that the classical guitar with nylon strings has.

It is not necessary to sample different models (the difference between those guitars that have a spruce soundboard instead of a cedar one may be important).

But the most important thing in terms of sound quality is that when you play a piece on the classical guitar you can produce different types of sound.

List the most important:
Related to the right hand:
1) typical touch of flamenco (toque apoyado): the finger that plays the string leans on the upper string
2) free fingernail touch: the standard way of arpeggio
3) fingertip; consists in not using the nail

    Related to the left hand

4) Legato: the finger of the left hand that begins the sound is changed with another (lower or higher in the handle) or ... also you start by playing the string empty and then you lowers a finger to change the frequency of vibration
5) Glissando: the finger of the left hand slides down or up on the handle while the string continues to vibrate

Then they can follow other features more common in soundfonts, but less important, such as stopped sounds, muted harmonics etc.

I am available to help, so when you want to contact me.

If there is a need for a more detailed outline, I can do it quickly.

In reply to by speedmeteor101

I can reproduce these sounds, but I have no experience on how to sample professionally.

I am equipped with a HP laptop with lots of memory and a good CPU.

If you know how to do it, explain it to me or give me a good link for this: I will try to organize myself by contacting a recording studio is needed.

I enclose the "instrument" file that I use;

I also tried to attach the AcousticGuitar.sf2 file but I couldn't; this file contains the sounds reported in instruments2.xml, sounds that I tried to create starting from public samples, which, however, were not really the ones I wrote you; in fact I am not satisfied with the result; if you want me to send you that file give me an e-mail address

NOTE: if you look at my instruments2 file you will notice that I use an "8-string guitar" to create my scores (so my basses that also come to A1 as in baritone guitars).

See you soon.

Attachment Size
instruments2.xml 2.59 KB

In reply to by gualtiero.chiapello

Honestly, I'm not the one to do send them to. Perhaps S. Chris Collins will see this and give you instructions... It's good that you have the instruments layout set up, though. That would make it easier to implement in theory. The way midi currently works, you'd probably have to set it up with multiple channels all using midi default sounds and then change the presets anyways if it became part of the release. Also about the sf2, it's not so much about the fact that an sf2 exists, but whether or not it's legal to include (the samples) in this larger soundfont.

In reply to by s.chriscollins

Hello

the samples I currently use are those of https://sites.google.com/site/soundfonts4u/
and precisely

Acoustic Guitars JNv2.4.sf2

where I made expressions associations with the following presets

           // preset nylon2


           // preset modified by me (see note 1)


              // unused guitar smooth preset - portamento required


             // preset modified by me (see note 2)


              // open muted preset


             //expressive nylon 2 presets

            // preset Spanish you can hear the sound of the soundboard

Note 1
I took preset 1 and removed attack and decay

Note 2
I took rol steel 1 (preset 9) and removed attack and decay and increased (a lot) sustain

It is obvious that I am not happy with what I have.

Do you still want to put the file on google drive or what I wrote is enough for you?

Tnx

In reply to by gualtiero.chiapello

Hi, I rewrote the excerpt from the instruments2 file because it wasn't clear

«Channel>
«program value="16"/> preset nylon2
«/Channel>

«Channel name="legato">
«program value="20"/> preset modificato da me (vedi nota 1)
«/Channel>
«Channel name="glissato">
«program value="1"/> preset guitar smooth non usato – necessario fare portamento
«/Channel>
«Channel name="armonico">
«program value="11"/> preset modificato da me (vedi nota 2)
«Channel>
«Channel name="muto">
preset open muted
«/Channel>
«Channel name="polpastrello">
«program value="17"/> preset nylon 2 espressive
«/Channel>
«Channel name="appoggiato">
«program value="22"/> preset Spanish si sente il rumore della tavola armonica
«/Channel>

In reply to by gualtiero.chiapello

... and finally in english version :-)))

"Channel>
«Program value =" 16 "/> // preset nylon2
'/ Channel>

«Channel name =" slur ">
«Program value =" 20 "/> // preset modified by me (see note 1)
'/ Channel>
«Channel name =" glissato ">
«Program value =" 1 "/> // preset guitar smooth not used - portamento required
'/ Channel>
«Channel name =" harmonic ">
«Program value =" 11 "/> //preset modified by me (see note 2)
"Channel>
«Channel name =" mute ">
// open muted preset
'/ Channel>
«Channel name =" fingertip ">
«Program value =" 17 "/> // expressive nylon 2 presets
'/ Channel>
«Channel name =" apoyado ">
            «Program value =" 22 "/> // preset Spanish you hear the sound of the soundboard
'/ Channel>

Note 1
I took preset 1 and removed attack and decay

Note 2
I took rol steel 1 (preset 9) and removed attack and decay
i increased (a lot) sustain

It is obvious that I am not happy with what I have.

Do you really want me to put the file on google drive?

In reply to by gualtiero.chiapello

Those unfortunately do not come under a suitable Open Source licence (ideally MIT, to match the rest of the soundfont) we’d require for use.

“The acoustic steel guitar was made by Keith Smith (www.keithsmith.ca) who graciously allowed me to use it.” is a restricted usage permit only for the person behind “Soundfonts 4U” and does not transfer to us. “Other sounds are the pick of the best that are freely available on the net” is not even a proper indication of source, let alone a licence. And the soundfont itself has “Copyright Information Not Present” and “Thanks to Keith Smith www.keithsmith.ca (SampleSmith.com) for the main samples used.
Presets and modifications by JN”. That’s it. Sorry, not suitable. (In fact, one probably cannot even legally do a derivative of it like you did.)

I am relatively new to MuseScore and have just completed a full arrangement for Gabriel’s Oboe for my Brass Band. I look forward to a new sound font for brass. I do not know how to create sound fonts, having access to all the instruments may be useful. Is it something I can create for each instrument? Good luck with the project.

How about a good tympani roll? Doing it with tremolo gives a weird effect sounding like repeated single beats, each one chopping off the decay envelope of the previous instead of blending with it.

Is this post still still up to date? @s.chriscollins are you still looking for samples? I could surely take some time to record some sax samples. The current ones are correct, but they really sound like jazz saxophone. If there happened to be an occasion to have both classical saxophone and jazz saxophone in the soundfont (like there is for brass, brass section, brass (expr.), etc.
A sax version without vibrato would be great too.

In reply to by Marr11317

For a while, Musescore was funding my work on the SoundFont, but this is no longer the case. However, I'd still like to work on improving it as time permits. I will eventually be creating a GitHub page for this project and actively seeking community involvement.

A while back, another forum member sent me a bunch of saxophone samples, which I tried to work up into the SoundFont. However, these samples proved problematic and didn't result in a musically satisfying instrument. If you would be able to record samples at both forte and mezzo dynamics, that should be plenty for me to work with. One reason the other sample set I received didn't work so well is because the forte samples were a hard, flat FF with no dynamic shape to them and no vibrato. What I would like to see from a new saxophone sample set would be for each note to be played as though it were integrated into real music, perhaps starting strong, and then settling slightly into a nice, natural vibrato. Be sure that each note speaks immediately (but not necessarily accented) to accommodate playing in fast passages. Recording each sample twice will help avoid bad notes. Lastly, you can share the recording session whole or cut (I don't mind doing the cutting), ideally as 24-bit or greater WAV, FLAC, or Audacity session files.

Thanks again for offering to do this! Peace :)

Yes, although some might not find those "improvements" helping. But in order to make Musescore's playback more appealing I believe we should have more options/soundfonts, so every composer can be happy. Now I'll share my own ideas about the sounds:
1. Slow and fast vibrato for all instruments that can do so: I'm not sure if we have that but just saying.
2. Less intense vibrato (for again all those instruments): imagine a sin/cos live on y and z axis, where y=pitch and z is just time... Let's say that the line is y=2sin(angle) ... I think it should be y=1sin(angle). In other words I believe that the current sounds look like, for example, a violinist moving his finger too much away from the pitch that is actually written in the score to create a vibrato, which is really intense. That's why I believe it should be less.
3. Stable dynamic in vibrato (again for those instruments): I just get the feeling that for some reason, when I hear vibrato, I also hear change in dynamic... if we were to make another sin() line, where y=dynamics, then I believe it should go from y=2sin to y=1sin just like above. I'm not sure if you get it... but I get it lol.
4. MORE MUTES! Not sure if strings have, ( if they haven't they should ) but I want to talk more about brass... especially trumpets. I think right now the trumpets have a mute soundfont that doesn't really sound realistic. So my point is: make it somehow sound more realistic (with and without vibrato).Also we should have more variation: instead of 1 mute, which idk what type is, we should have at least Harmon and Straight mute (and other common ones that I don't mention perhaps).
5. When an instrument starts doing vibrato, it should start slowly and then go faster... I think most will find it more "professional-sounding".
6? BONUS:optional (actually everything is optional because it's just sugestions)... the sounds should sound more "realistic" in matter of timber (timbre?) .Violins sound a bit *dead like flies or something. Trumpets on the others hand sound too intense : they should sound a bit more lyrical (well actually that's for the with-vibrato sound). Well those 2 for now.
6. I find this one important: I think sometims that som instruments, especially strings' sounds, start piano and do a crescendo when the sound starts. Just imagine a melody, it sounds like: d o oo oOoOOO r e eeEEEEE m mi iIiIII... I find that it would be better (not perfect, better) if it sounded like DO RE MI. In other words, have a "straight", in dynamic, sound for each instrument.
Advice: I think if the sounds are gonna change, then we should keep some old ones that some composers might prefer.
Okay that's all.

In reply to by Iothes

P.S. Sorry if some of those have already been suggested (did not read those 2 pages in this post).
Also I believe it would be good to look at Sibelius' sounds, as there are many who say they are great... Just saying that those sounds could have some nice characteristics that Musescore doesn't have. But I gotta say that Musescore's sounds are great too.

In reply to by Iothes

lothes, some observations. I could be wrong, but it sounds to me that many of the free fonts out there are based on a few "master" fonts. AFAIK, no notation software offers any of the things you suggest. The closest I know of is the choice to have vibrato or not.

  1. I believe that even Sibelius only offers a straight mute. Some one will have to record a player using the different mutes and find a way to match sounds already in a particular font, and within GM limitations. For free. People are making their own fonts. But it is not for the faint of heart, so to speak.

  2. This really depends on the music. Perhaps on certain long notes. But not all.

  3. I find this to be true with some fonts. I don't use them for that reason.

Consider that Sibelius sounds are some 35 GB of professionally recorded and manipulated sounds. And not GM bound like MuseScore. Free fonts are typically less than 1 GB. Let's face it, free fonts cannot be as good, and are not intended to be.
Some of the things you suggest are not up to the composer. Or the software.
Personally, I'm impressed with how far MuseScore has come. But there are limitations that can't be helped.

In reply to by bobjp

I was gonna suggest sound recording, and because I know it's costly (for some reason I don't understand) I would even donate, just like VSCO's sounds ... that you pay actually. I want to die... :') but to be honest... I think that it's possible to fit them in less space. But well it doesn't matter now.

In reply to by Iothes

The base SoundFont is not going to be able to contain every known articulation and playing style, but my goal is for it to eventually have a better set of defaults. Eventually, if MuseScore supports VST plugins in the future, you can go out and buy that 500 GB orchestral library to maximize playback realism. Of course, no matter what instruments are replaced, there will be those who prefer the old versions, so I would like to also create a SoundFont containing all old preset versions so people can still access them easily. I will create this once more work is officially under way on replacing instruments in the core SoundFont.

In reply to by dhfx

See https://en.wikipedia.org/wiki/Virtual_Studio_Technology
Basically, VST is a plugin format that can be used to manipulate complex sound data.
For example, this means that you could have a single violin sample, and use VST to create, out of that single sound, a vibrato sample, or an ensemble sample, by manipulating the audio programmatically. This reduces drastically the number of samples that have to be recorded and downloaded.

In reply to by Marr11317

Still lots of work, and partially very platform-specific.

MuseScore is, first and foremost, notation software; it aims at good enough sound playback without going into all these specifics or becoming a full DAW. The core developers aren’t going to work on that; getting the notation software itself to produce beautiful sheet music is more than enough effort already ☻

In reply to by mirabilos

Jumping in here to add clarify a couple of points from the perspective of MuseScore as the company.

It is absolutely accurate to say that MuseScore is notation software and the current playback provides reference to the composer in their primary task to create high-quality and beautifully accurate sheet music for performers.

But...

The future of MuseScore is not simply notation software, but is instead composition software that supports the workflows and goals of the modern composer.

For the modern composer high-quality and beautifully accurate playback is of equal importance to high-quality and beautifully accurate sheet music.

In order to accomplish this, full VST support and significantly improved MIDI capabilities are essential.

This will take considerable time and effort, and the timeline for when this might happen is not yet determined.

What is clear:

1) High-quality playback is as important as notation capabilities
2) Improvements to playback cannot come at the cost of reduced progress on notation
3) We are committed to making this happen

In reply to by Daniel

"For the modern composer high-quality and beautifully accurate playback is of equal importance to high-quality and beautifully accurate sheet music."
When writing or arranging music for live performers, they will be making music from their own interpretation of the symbols/graphing/notation you have on the page and their understanding of playing music on their instrument, right? Furthermore, there are different styles and genres that have those intrinsic characteristics and rules that aren't ever written on the page. Accurate playback can mean many different things for different types of styles of playing, so from that standpoint alone, this seems somewhat unrealistic; I'd think that if you have an idea of what you want in your head, having reasonably good playback can give you some reassurance on some decisions rather than relying on playback to provide choices or something.

"The future of MuseScore is not simply notation software, but is instead composition software that supports the workflows and goals of the modern composer."
As a happy user, I'd say it already does this quite well; if I'm making sheet music, I know who I'm writing it for and give them exactly what they need to know to create a sound close enough to what is in my mind. If, by modern composer you mean creating soundtracks and whatnot, I prefer notation too, so I first write music in musescore, export the midi and import it into FL Studio to have 100% control over the playback and sound.
I guess for me, the playback has always been good enough

BUT - I did have a suggestion for capturing different styles (that would not affect notation). Since the piano roll allows the user to have full control over the attack, release, length, and velocity, regardless of the notation, what if there were downloadable files that the user could apply to a score to automatically change those aspects of playback to reflect a desired style [using the piano roll]? (Like you press an 'apply style to score' button or something) I'd imagine a menu could be used to make these files, or at least a basic setup similar to the instruments xml file (but hopefully more intuitive). Then you can just add lines of actions for musescore to apply to the midi; I'd think it would work smoothly if it was made with a bunch of 'if then's. For example 'If this symbol [♪.] is present [+50] to [On Time], [+80] to [Velocity Offset], [+100] to length. Or if [example rhythmic passage, certain note length, etc.] comes before/after this symbol [♪.] or if there is a 'crescendo' attached to this symbol etc... And then if musescore detects any possible overlap the user would have to set priority. Then, even if the sounds themselves aren't the most realistic the user can still get a more realistic interpretation of the notes by playing with a relevant style.

I do think it would be cool to have more realistic playback, but I haven't even seen VSTs do it that well by themselves yet, so I haven't set expectations for anything. Still hoping for any improvements though. It's a shame that this project is not being funded anymore... can we kickstart it, maybe?

In reply to by speedmeteor101

You are correct in that the interpretation aspect of playback is not implicit in VST support.
Supporting VST is simply the starting point.

To create more accurate interpretation it should be as you have described, easily customizable within the piano roll editor (which should also have equal priority to notation).

You should not need to export MIDI from MuseScore and import into a DAW to get the desired result. All of this should be possible within MuseScore.

Regarding funding of the project, in addition to the generous and invaluable efforts of the community there are several full time contributors to the project that are employed by MuseScore, which is owned by Ultimate Guitar. Considerable investment is being made in accelerated development efforts.

In reply to by s.chriscollins

I think a better solution to relying on musescore to provide better quality (whatever that means) playback is to improve its MIDI out capabilities. It's far better to load up a bunch of sample libraries in an external host, such as Carla or Ardour, and route Musescore's MIDI playback to the host's MIDI-in rather than relying on Musescore to be a plugin host too.

This has several advantages:
* It eases the development burden on the Musescore team. Just add more MIDI-out functionality.
* The support burden on the Musescore team will not need to expand to accommodate issues created by the vast range of VST plugins available.
* It removes the limitation of a single plugin architecture (VST) and allows users to use whatever plugins they prefer, aax, au, LV2, LADSPA, etc.
* Switching between Musescore projects or recovering from a crash won't require every sample library/plugin that you're using (possibly several hundred GB) to be reloaded.
* Plugins can be hosted on a separate computer and interfaced via LAN.
* Plugins will be isolated limiting the chance of an unstable plugin crashing Musescore.
* If, like me, you switch between projects running in different DAWs as well as Musescore and want to share the same plugins between then, without reloading each time, the external host system enables that.

In reply to by reddiesel41264

Unfortunately, MIDI isn’t really a first-class target in MuseScore either. It’s a side product of MuseScore driving one of its two (optionally three) internal synthesisers. (See the MIDI track 0 issue.)

Do also keep in mind that not everyone has, or even wants to, install the full audio workshop chain. For example, with what I do (mostly SATB + accompaniment), the built-in playback is completely sufficient. This should not worsen.

But yes, in general, making the MIDI output more helpful for integration with other tools (be it DAWs or things like Hauptwerk) is probably useful enough in general. This needs people experienced with those tools, I guess.

Well, for me my main concern is the timbre. Also, right now I was writing the theme of the 17th Prelude of the First Well Tempered Clavier in a violin, and it sounded bad because the violin's note started piano going to forte midway, it was like: the violin played only the second halfs of the notes (or you could hear that). That's why I think all instruments with similar case should have a soundfont that lets them start "straight" and not with crescendo. If you know what I mean.

In reply to by Iothes

I will eventually be creating a GitHub page for the SoundFont project, once I finish developing my Python build system for it. This system will enable building both versions of the SoundFont from source (presets, instruments and samples will be broken down), and will facilitate greater collaboration as well. It will also provide a place where the latest version can always be downloaded and beta versions can be tested.

I also want to state that the dynamics need to change for every instrument... This sentence alone doesn't explain what I want so here:
Have you noticed that if you have 20 instruments play a chord in the same time AND same dynamic then some instruments will sound more than others while some not even a bit, like they do not exist? Trumpets for example sound a LOT and cover the other instruments. While some woodwinds and strings won't sound if there are brass playing at the same dynamic as them.
I know that this happens in reality too, but I think not so much.
But please also consider making the timpani sound more majestic in forte dynamics (and also add straight and Harmon mutes for brass... Just those two please ).
Also... This might be too ambitious but what if Musescore had some system LIKE NotePerformer? ( But for Mac and Linux too)... Yes this idea might be impossible in the meantime.

In reply to by Iothes

I've found that for multiple instruments playing together, I often have to individualize the dynamic markings in order to keep some instruments from drowning out others. Each instrument seems to have its own dynamic range which is some function of the velocity, but not necessarily equal or proportional to it. You can also set the level for each instrument in the mixer to compensate for the differences.

In reply to by dhfx

Yes those solutions take a while for big orchestras... Also I tried modifying the mixer and not much changed. But I think the fact that brass generally sound more than other instruments... That just seems like the case in most recordings. But what I think should be done in Musescore is to have every instrument in the same level.

In reply to by Iothes

I disagree strongly and imagine most people writing for real musicians would too. Trombone are louder than bassoons, that’s just how it is. People writing for these instruments need to be aware of this and keep it in mind and write accordingly. MuseScore needs to accurate reflect how the music will actually sound once played for real.

In reply to by Marc Sabatella

Oh ok... It's just that recently I have been exploring new recordings and in those new ones (probably 2015+) the strings, flutes and clarinets seem to sound as loudly as some brass. Oboes and bassoons (and double basses) seemed to sound less.... I just don't like how the current brass sound A LOT louder than the other instruments and oboes Do not even sound a bit (and strings too I think).

In reply to by Iothes

Modern recordings are often "striped", that is the instruments are recorded separately and then mixed together afterwards. This is typical for modern film scores. This usually results in a totally unrealistic (although artistically pleasing) balance between different instruments.

When writing a score for musicians to play together in the same room you have to do your mixing in the score, with orchestration. When it comes to the recording the conductor and section leaders will also help to refine the balance, but (for example) a trumpet playing forte will always be louder than pizzicato strings playing forte.

It's good to remember that although we think of dynamic markings ranging from quiet ppp to loud fff, the words piano and forte really mean soft and strong, they are not literally indications of volume. So telling a trumpet player to play "strong" is different to asking a string player to play "strong".

Other things affect volume, including the range the instrument is playing in, wind instruments especially have a great variation in their volume across their range, and individual players also have their own range limitations.

In reply to by reddiesel41264

>0< I GET IT ! I just stated in my last comment that some of those things that sound in Musescore feel to edgy: oboes are too piano sounding while trumpets seem to play Too loud in the same dynamic as oboe. Other instruments might have similar cases. Scrap the idea of having every instrument have the same loudness. But things as the oboe and the trumpet sound too unbalanced.

In reply to by Marc Sabatella

I agree with you on the question of brass vs woodwind volumes, but trumpets and trombones in unison do tend to sound quite harsh in timbre (as opposed to simply being "too loud"). My bigger issue within the brass sounds is how quiet the French Horns sound in comparison to trumpets and trombones. Also, I'm a bit offended by the lack of Euphonium and Baritone representation (in the soundfont) :P

In reply to by osubuki_

Ye, I also find cellos and contrabasses to be a bit loud to (at least in the rite of spring for Musescore). *I generally believe that Rite of spring, whic can be found in musescore.com is good for testing (probably not everything that is dynamics-related can be based on this but other characteristics of the sounds heard) By the way I saw, in the same score, that when there are chords which must sound accented, violins always play downbow (or have a shape above their notes which look like Π). Does that means downbow should sound a bit like accenting (lol this word)?

In reply to by Iothes

It really depends on what you are trying to produce:

  1. If you are trying to compose a piece for a live group, you have to already have the knowledge of how those instrument work together, and mark parts accordingly. That's exactly what notation software is meant to do. Even DAW users (who can reproduce and orchestra far better than anyone with any notation software) say that there is no substitute for the real thing. If it could be done, it would put real musicians out of a job.

  2. If you are trying to reproduce a recording, consider that any recording is only as good as the recording engineer's take on the performance. In which case you should mark dynamics however you need them to sound the way you want.

  3. Trying to make an mp3 sound like a live performance seems to be the strangest of all. If you are sitting on one side of the orchestra, you may not hear some instruments on the other side quite as well. How often do we get to sit in the exact center in an acoustically perfect room?

You can't learn orchestration using notation software. Why would you want to? Music is meant to be performed live. Software will give us a rough idea of what to expect. But remember that out in the real world, it might not matter how you mark something. The musicians will work their magic on the part. That's their job.

In reply to by bobjp

  1. Yes, 2. That last part... if that happens it cannot be helped. Why not have the capability to compose and record without needing people? (I'm also a pianist and I know what you think but then again, it's an advancement). 3. Also yes, but not much will change if you sit in other places of the same room (unless the room is not designed for acoustic performances).
    I agree with the last part, and based on 3. I believe that it could give a rough idea, so the new soundfonts should try to sound like an orchestra in a room meant for it. By the way, if you are talking about dynamics, I already have quit that idea... we are just examining things on that part that sound a bit too "edgy" (lol I used the same word again)

In reply to by Iothes

If you are using the computer as an instrument, that's is certainly one way to look at it. The problem comes with the same limitations and sound file runs into. Even if you manipulate the poor sounding trumpet font, there is no way to know it will sound good on all systems. For example, I mix something that sounds great through my Sennheisers. It will sound much different on my car stereo. And even more different on the sound system at church. When I wrote music for a play at the church I had to remix everything to sound the way I wanted on that system. This is what a conductor does with a live group. Recording engineers have special protocols and equipment that produce recordings that sound good on the most varied equipment possible. You and I don't have that access, nor the knowledge to use it.
The vast majority of concerts I have attended or played in, were not in rooms intended for orchestra concerts. Even in a multi-million concert hall, it makes a difference where you sit. As a trumpet player, I have been in the midst of a performing orchestra. I have been exposed to how instruments blend together from the inside out. But I can't hear everything. That's why the conductor is so vital. Especially in a new situation.
To be a conductor, or musician, or recording engineer, takes decades of work and an expensive education and special (read expensive) equipment.

In reply to by Iothes

Sorry, but by nature, different instruments have different loudness with the same dynamic marks (p, mf, f...). Composers are aware of this. Conductors also adjust those levels, and in Musescore, it is the mixer window's role. Same levels must not be applied to all instruments in the soundfont. Or else precise well-written scores will sound bad.

In reply to by Iothes

Wow, this whole thread is a lot of text for just "use the mixer".

Different composers, different conductors, different performers have different concepts of balance and blend.

Just use the mixer and find your sound.

Also, the default soundfont will be completely overhauled for MuseScore 4 - from absolute zero.

So, there is no need to get into minutiae on current soundfont (which has been a major leap ahead thanks to Chris Collins) because none of the same samples or same source will be used and it will be entirely rebuilt from the ground up.

In reply to by Daniel

Sorry you are right. I happen to mess with the mixer but never notice any significant changes with the instruments I find too strong or weak sounding... like oboe yet again. And I just wanted to express my opinion about those instrument which, I will say again in my opinion, sound too strong or too weak (by the default settings)

I’ve been supporting a user on IRC who’s on a Raspberry Pi 2. This device is seriously underpowered, but with the MS_General soundfont we’re hitting RAM issues (it has less than 1 GiB). I’ve directed them at timgm6mb for now, but they liked the new default sound.

Perhaps being able to subset the font, so as to have only one or two instruments, would help the RAM issues?

Other than that, of course, a code change to the synthesiser to load only the parts of the soundfont to RAM that it actually needs (or mmap the file, even… trickier with SF3, but even there it doesn’t need to Vorbis-decompress the sounds for unused instruments IMHO) would help, but I’ve given up holding my breath for that (because it’s the thing that’d make sense)…

In reply to by mirabilos

What is the desired RAM footprint for the uncompressed SoundFont on Raspberry Pi 2? The system I will be putting together will allow custom-building SoundFonts for MuseScore and can include various subsets as needed. Furthermore, GeneralUser GS will be updated to support single note dynamics and will be available to the build system as a low-RAM option (it's currently 29.8 MB in size).

In reply to by s.chriscollins

I guess it depends. I’d really prefer MuseScore to not keep it all in RAM as primary option.

For most systems I’d say 512 MiB (which is slightly more than even the current HQ soundfont) is a good cutoff, but for the Raspberry Pi I’m trying to stay at or below 64 MiB. This is a guesstimate.

GeneralUser GS is nōn-free, and as such I cannot recommend it to anyone, rather the inverse. (Sorry. No slight against you, but we need accountability and licences for all sample sources.)

Hi
Sorry i'm not a expert, and I also don' t have the necessary audio equipment to check it, but looks all or most of the drumkits presets are in mono. I know the soundfont is based on Fluid (R3) Mono GM and I know they are a (Original Stereo version) Fluid (R3) GM SoundFont (MIT) .
I really don't know if they exist a license problem.
I really think that a stereo drums sounds really more natural, and I think could be a "easy" improvement for the soundfont.
Thanks

Josep

Do you still have an unanswered question? Please log in first to post your question.