Uncompressed version of amended Fluid soundfont
I understand Michael (ChurchOrganist) made adjustments to correct issues in the original SF2, but is this available in SF2 form (if so, apologies)? It would be ideal for the stereo and uncompressed qualities - both for the musescore.com server in producing the MP3 version (avoiding double compression?), and for offline use.
The sound on .com might be affected by this, so I attach an mscz (produced in 2.0 RC on Mac 10.7.5) and the MP3 that was downloadable to demonstrate.
| Attachment | Size |
|---|---|
| Hi-hat.mscz | 5.91 KB |
| Hi-hat.zip | 84.62 KB |
Comments
The uncompressed SF2 version of FluidR3Mono is always available from the Soundfonts forum. CUrrently from my DropBox account.
To save you searching - here is the link.......
https://www.dropbox.com/s/soi5tj3s3113f26/FluidR3Mono_GM_sf2.zip?dl=0
I'm not sure what you mean by double compression?? The Ogg compression of the SF3 version has absolutely no effect on the MP3 compression of the audio file.
The original FluidR3 soundfont is still available from various download sites. I suggest you use Google to find them.
In reply to The uncompressed SF2 version by ChurchOrganist
You're quite right about the FluidR3Mono being available in SF2 form - I had thought it was SF3.
By "double compression", I was questioning if the MuseScore server was using a SF3, then making a MP3 from this (I think it should use the SF2 on the server to produce the MP3, if it isn't already).
I have the original FluidR3.
In reply to You're quite right about the by chen lung
Presumable the musescore.com server is making audio from the the SF3 (if it's not using SF2 - you'll need to ask Thomas about that).
The server will then compress it for an MP3 file using the MP3 compression algorithm whether the source soundfont is sf2 or sf3 is completely irrelevant as the MP3 is generated from produced audio - not direct from the soundfont.
In reply to Presumable the musescore.com by ChurchOrganist
MuseScore.com backend is currently using the SF3 version included with MuseScore 2.0 RC and export to MP3 in the same way that if you do it via File > Export.
If you actually experience any problem with this, please tell me.
There is a bit of difference between a WAV and MP3 exported from 2.0 RC (see attached).
The MP3 downloaded from musescore.com (see first post) sounds worse than either of these (mushy?). Is it because of different settings (as mentioned here )?
I think, ultimately, the server should use SF2 (to produce the audio, then compress it to MP3?) so the maximum quality is retained - especially if users opt for the SF2 (although they may have a problem doing this, currently: #21592: Exported audio doesn't play loaded soundfont).
In reply to There is a bit of difference by chen lung
There is *always* difference between WAV and MP3 if you listed hard enough. MP3 is compressed, WAV isn't, and they both have lots of different quality settings available. The only reason to ever use MP3 rather than Mp3 is to save space, so we apparently decide to take that seriously and save a lot of space by using high compression / low quality. If you prefer high quality, use WAV. If you prefer some sort of compromise, use WAV then convert to MP3 using third party software that gives you control over the settings.
In reply to There is *always* difference by Marc Sabatella
Yes, I expect quality loss between both formats (I just thought the difference was a bit suspicious), and I don't mind the lack of control as such within MuseScore itself (I'd export the WAV to an audio program) - my problem is, the MP3 from .com sounds worse than the MP3 exported from the desktop software, and that is what will likely used on the website, and the app.
In reply to There is a bit of difference by chen lung
There is *always* difference between WAV and MP3 if you listed hard enough. MP3 is compressed, WAV isn't, and they both have lots of different quality settings available. The only reason to ever use MP3 rather than WAV is to save space, so we apparently decide to take that seriously and save a lot of space by using high compression / low quality. If you prefer high quality, use WAV. If you prefer some sort of compromise, use WAV then convert to MP3 using third party software that gives you control over the settings.
In reply to There is *always* difference by Marc Sabatella
Marc it is clear that you do not fully understand about MP3 compression.
There is a difference between compressing audio to an acceptable quality and compressing it to the extent where it introduces hiss and distortion which is the current case with MuseScore.com
Throughout it's history 128kb/s has been the default quality for MP3 as most people cannot hear the difference between MP3 audio compressed at 128kb/s and a full PCM waveform. Some people, usually musicians, can detect artifacts in the high frequency element of the 128kb/s MP3 audio, but in the main this is not the case.
Compressing any lower than that successively introduces more artifacts recognisable by most people, and by the time you get to 64kb/s which is the current setting for musescore.com you have a quality equivalent to FM radio which contains hiss and other distortion readily audible to most people.
Given that MP3 compression at 128kb/s results in a file size around 1 tenth of the original PCM file it seems silly to compromise audio quality to that extent purely for the sake of a smaller file size.
OK you can download as WAV and compress it yourself, but how many normal users would know how to do that? OTOH the defects in 64kb/s MP3 audio are so glaring that they will give musescore.com a bad reputation for audio quality it does not deserve.
In reply to Marc it is clear that you do by ChurchOrganist
I understand the *concept* just fine; although it is true I have no no feel for the actual sound of the different levels. I normally use WAV or MP3 256 or higher when I care about quality because I can hear the difference; levels below that are all about file size to me.
So I'm not saying I think 64 is better / worse than 128, just that *apparently* (the word I used) a decision was made to agressively favor small file size over quality. I never said I agreed or disagreed with the choice.
In reply to Marc it is clear that you do by ChurchOrganist
Let's assume we all know what we are talking about... but we just lack information on what's going on in MuseScore, MuseScore.com and fluid.
The question of the OP is: Is there any indication that the SF3 soundfont + MP3 encoding is a problem?
If there is, I'm not aware of it. The OP use "might" and is full of precaution. Is there a problem here? I don't think so. If yes, please give an example, an mscz file and an MP3 and we will see how we can fix it.
Next, currently MuseScore exports MP3 at 64kb/s and 44.1KHz. It seems to be a problem and it makes MP3 export unusable for some people. For other people, it's not enough compression... and the file size are still too big. The ultimate solution would be to be able to change the compression settings but we can't do that in 2.0. It will have to wait. As a compromise to the audiophiles camp, I will put back the default to 128kbps/44.1kHz.
Next MuseScore.com. For (wrong) historical reason, for lower bandwith consumption etc.. MuseScore.com has *always* exported MP3 in 64kbps/24KHz.
24kHZ is not even a standard and I can't track why we (in fact I) decided to do this... but it's a fact. Nevertheless, nobody never complain about this being too low quality. The playback on musescore.com has been criticized for many reason (lack of realism, too mechanic, too much or not enough reverb etc...) but it has never been criticized because it's too much compressed and so have artefact.
Currently MuseScore.com uses the 2.0RC default for MP3 export (64kb/s and 44.1KHz). I'll change ASAP to go back to the previous compression (64kbps/24KHz) since apparently everybody was happy with it. If it's not the case, please let me know! In particular, ChurchOrganist, can you hear the "so glaring defects of 64kb/s MP3 audio" on musescore.com before the introduction of MuseScore 2.0?
In reply to Let's assume we all know what by [DELETED] 5
Maybe nobody has complained about MuseScore.com's MP3s' quality, because up to recently it used Tim6MB (didn't it?) which was bad enough already, but also small enough to not produce artifacts?
In reply to Let's assume we all know what by [DELETED] 5
TBH audio download was not something I used in either .WAV or .MP3 format before, and probably will not do so again unless for testing purposes.
I am however, linking to a zip file containing audio generated from my latest upload to MuseScore.com (it's too big to attach)....
https://musescore.com/churchorganist/scores/708191
The zip file is here......
https://drive.google.com/open?id=0B7cZM0RQwkwSR05zQnlsYWdvczA&authuser=0
The 64kbps mp3 is downloaded from MuseScore.com
The WAV was generated from MuseScore 2 RC
The 128kbps mp3 was generated from the WAV.
On doing a comparison listening test it can be heard how much grittier the oboe sounds in the 64kbps mp3 file this is typical of MP3 compression which at 64kbps is recalculating a high proportion of frequencies - this article from SoS explains it in quite some depth if you are interested http://www.soundonsound.com/sos/may00/articles/mp3.htm
It seems to be high frequency band components which suffer the most from this recalculation - to me there is a clear difference in the sound of the oboe between the pure PCM (WAV) file and the 128kpbs mp3.
I leave you to draw your own conclusions.
In reply to TBH audio download was not by ChurchOrganist
As I said earlier, the whole discussion (which is not really at the good place, since the OP is not asking anything about the parameters of MP3 compression, but anyway) made me rethink about the purpose of MP3 export in MuseScore. So I will change the default as follow, comments are welcome.
Before (MuseScore 2.0RC)
- CBR
- 64kb/s
- 44.1kHZ
- Joint Stereo
After (so for MuseScore 2.0)
- VBR
- target ~190kb/s
- 44.1kHZ
- Joint Stereo
FWIW, the OGG export uses 44100 Hz, ~128 bps.
In reply to As I said earlier, the whole by [DELETED] 5
I'm not sure VBR is a good idea, ISTR that not all MP3 player support it?
In reply to I'm not sure VBR is a good by Jojo-Schmitz
While that may have been true in the past Jojo, I think that position may well have changed during the past few years.
Certainly Microsoft and Apple players support it.
In reply to As I said earlier, the whole by [DELETED] 5
FWIW, that's more consistent with what I might normally choose. As I have said, I don't really play with this much, as WAV is more normally what I want anyhow - and if I do use MP3, it's usually higher quality still. But I do distribute MP3 to other musicians and when I have done tests, I have found 192 is as low as I am really comfortable with. Normally I use CBR just because VBR is still "new" to me, but I gather it should give comparable quality at a smaller size.
Meaning your proposal does like a good compromise to me personally, although I recognize everyone will have different preferences. If 192 VBR can come within a factor of 2 or so of the old sizes (and a quick test shows that's about right), I think it will be good.
In reply to FWIW, that's more consistent by Marc Sabatella
Thank you for your input!
I read a bit about VBR and compatibility issues. Apparently, it has been solved in the past years indeed, but of course, mainly for digital players... There are still lots of car audio systems, mp3 players in the wild which do not support VBR correctly...
I checked the default export settings in Audacity and they are 128kb/s - CBR. Also iTunes on mac, export in CBR by default but 160kb/s. It's their high quality settings. They have a "good quality" setting at 128kb/s.
So I might be using CBR 128kb/s ....
In reply to Thank you for your input! I by [DELETED] 5
I've been using VBR for all of my MP3 needs for years and I have not ever come across a player that couldn't handle it. Maybe 10 years ago that would have been an issue, today its a moot point.
I haven't really been following this discussion completely, but if MuseScore has an MP3 export and it has to be one setting, I would probably opt for either CBR 192 or VBR[0] as the setting, which is a reasonable amount of compression and truly you will need an audiophile system to hear the difference compared to WAV. It ends up around 1.5MB per minute, give or take, depending on the content.
In reply to As I said earlier, the whole by [DELETED] 5
I haven't noticed a commit with these changes, and musescore.com doesn't seem to sound different from before either.
For simpler access, here are the links (already-posted in this topic) for comparison:
mscz
musescore.com
MuseScore 2.0 RC
Michael, do you intend to produce a stereo uncompressed version of your modified soundfont?
In reply to Michael, do you intend to by chen lung
Why would you want one?
If you think about it most real instruments are monophonic sound sources, so producing a stereo saxophone (for example) in a soundfont is nonsense.
FluidR3Mono is so called because the instruments (ie the samples) have been converted to monophonic sound sources. It is still outputting stereo audio, and you can move the instruments around in the stereo soundstage.
The interesting thing about the original Fluid soundfont was that the samples you might expect to be stereo, such as synths and pianos are not - they are mono samples. The samples you would expect to be mono like sax, violin, guitar are stereo. So it was not built on good sound design principles.
In reply to Why would you want one? If by ChurchOrganist
It was indeed for the instruments you mention, but thanks for explaining.
I'm reaDING AND READING -understand less and less.I downloaded 2.o again this morning.Like the new setup. Wanted to put some of the other sf 's in synthesizer .This is easy to do but I've been using fluid for 3 years now and the sound was so good I think somehow it magically decompressed by itself as unlikely as that seemed.I have the sfark software but am not sure how to decompress within its own window.Ofcourse I want to use the new sf of sonatina but am so confused .I have a few hours to figure this out.I know most of the users are 1/3 my age so there must be a video on youtube explaining or showing this:I will now go and find.I love the work your team has done and the lifeline it has provided me for a number of years now.I really just wanted to say thankyou again !
In reply to I'm reaDING AND READING by 21st CenturyBoi
If you have sfArk installed on your machine you should simply be able to double-click the .sfArk file and it will aitoagically decompress into an .SF2 file.
HTH