The SoundFont 2.0 Specification!!!
It is available for free (smiling face) here: http://connect.creativelabs.com/developer/SoundFont/Forms/AllItems.aspx
Again, here: http://connect.creativelabs.com/developer/SoundFont/Forms/AllItems.aspx
Do you still have an unanswered question? Please log in first to post your question.
Comments
Thanks for posting this - it was on my list of things to do :)
There are other useful SoundFont resources which need posting here to.
I will get round to that myself eventually, but if others wish to post them first - please feel free :)
In reply to Thanks for posting this - it by ChurchOrganist
I don't know where the SoundFont 2.4 specification is, but I think the only difference between SoundFont 2.1 and SoundFont 2.4 is that 2.4 is 24-bit, while 2.1 is 16-bit (because the two versions are cross-compatible - all 2.1 files work in 2.4 and vice versa). Do you think so? What are further differences?
In reply to SoundFont 2.4 by Musical Notation
You can get the 2.4 spec here:
http://freepats.zenvoid.org/sf2/
As far as I'm aware the only difference is 24bit sample support.
I've arrived here via John Kirk who has suggested creating a British Brass Band Soundfont. I'm interested in contributing to the project but there seems to be a lack of specification details here.
In my University studies I had to create a Xylophone Soundfont using Apple Logic's EXS24 sampler and for my MSc i created an app for demonstrating how digital filters work. During this project I used the Windows XP Roland DLS (downloadable soundfont ) as a source for filtering. www.psybersonic.com Not legal but it's old and I was using for educational and non profit purposes. You can download the app from the website - needs Java but runs on Mac and windoze.
This app gave me more insight into how soundfonts work so I have a few questions before I consider organising recording sessions.
Which formats does Musescore accept? DLS is not very common so are we talking Creative Labs here?
How many volume levels are we considering?
How many articulations?
How many sounds need to be recorded per octave? In the case of brass this could be per harmonic, is there any research on this?
These are the the elements that we need to establish before we go into a studio. I would try and get some time in an acoustically neutral booth at the University of Kent although I am no longer a student there.
In reply to Soundfont by psybersonic
First of welcome!
To answer your questions:-
The base around which we are working is the Fluid synth which recognises the SF2 format and is the primary synth engine for MuseScore at present.
If you are considering submitting a complete soundfont then SF2 compressed with SfArk is the preferred format.
If on the other hand you are submitting samples then they shouiold be submitted in 16bit 44.1kHz Mono format - preferably using FLAC compression, but raw PCM would be OK.
Regarding the volume levels, articulations and sounds per octave, then this will vary according to the instruments being sampled - in other words I will (not being a brass player myself apart from 6 months of French Horn lessons in the dim and distant past) accept whatever Jon and you think is necessary to produce a top quality Brass Band soundfont - but you may either have to educate me as to how this would be put together, or produce a the soundfont yourselves.
Hope this is not too vague.
Basically we're looking at producing as high a quality result as possible - MuseScore 2.0 has the ability to assign multiple Soundfonts to a score, so we can concentrate on quality rqahter than cramming the entire GM soundset into a small space.
If you have any further questions please come back, but I think we probably need a new thread for this, so please open one entitled Brass Band Soundfont or similar :)
Regards
Michael
In reply to First of welcome! To answer by ChurchOrganist
But here is where I think having general guidelines as to what we are trying to achieve would still help. I don't think it makes sense to have 100MB worth of trumpet samples covering a variety of velocity and articulations, sampled on each and every note, and then only 20K worth of flute samples, with just one velocity and two sample points.
The idea of multiple SoundFonts would presumably be so we could group the samples into smaller individual sets that could be loaded as necessary, so we don't end up with one enormous 50GB soundfont that most computers couldn't handle? If so, then some discussion of that grouping also seems in order.