how to make my own soundfont and distinguish between different attacks with musescore?
I know how in musescore I can select between pizz & trememlo & regular for strings on the default soundfont. If I want to make my own soundfont of me recording my saxophones for all different styles of attack, then what is the best format to use and how would I differentiate between different types of attacks. I notice in the default soundfont, if I put an > accent, then it simply increases the volume volume by ~32 midi levels, while if I put a staccato, it shortens the duration by approx 1/2, which may be a valid interpretation of those articulations, but I would want to record all my notes with all different types of attacks. Also I think the sfz format permits recording different samples for different volume levels, so I would want to record for all main dynamic levels. I'm reading what I can quickly on the internet and have been looking at http://www.sfzformat.com/index.php?title=Opcode_Types but am not quite sure how to go about creating such a soundfont that would work in musescore. Are there any FOSS soundfont editors are there that would work well for creating such a soundfont for musescore?
Comments
You cannot create a Soundfont like this to work with MuseScore unless you want to create a special instruments list for the program to use.
There's Polyphone.....
http://www.polyphone.fr/
Like the SFZ format, SF2 soundfonts allow the creation of velocity splits to allow different samples to be used with different volumes.
It would also be possible to have different attacks, but you may have to set up a special Instruments.xml definition for your saxophone in order to place the different attacks on different channels.
Basically you would assign the various attack sample sets to different banks within the Sax patch number (Tenor is #67). You could then assign channels to each attack for use within a score.
It could also be done with the Change Instrument facility in MuseScore 2
In reply to There's by ChurchOrganist
Thanks, this is exactly what I need!
In reply to There's by ChurchOrganist
Incidentally I think that MuseScore would default to using Bank 0 for each patch number if the defined bank were missing in the soundfont (haven't tested this so I'm making a bit of an assumption here based on what I know of the existing sound engine).
This would mean that such modifications to the Instruments.xml file could be used with the default soundfont, but only activate the attack banks when the correct soundfont was loaded.
So if you're prepared to share your work with the rest of the MuseScore community, that would be a great step forward for improvements in Sax sounds.
In reply to Incidentally I think that by ChurchOrganist
Thanks for your help. So far I've only done tenor saxophone with regular attack at medium volume level. I recorded all notes from Ab2 to Eb6 (which includes an octave of altissimo above regular range), and imported into polyphone, and used the "autoloop" feature (which seems to work ok for the most part). Here is example playing reunion (notes in red are beyond my range I recorded on tenor):
https://www.youtube.com/watch?v=mgKPAkUIXG8
download current state of my soundfont
I intend to make soundfonts for all saxes I have: bass, bari, tenor, alto, soprano, sopranino, and maybe with my flutes and clarinets. And as I said earlier I will do with different volume levels and different attacks. Also, I have many different mouthpiece/reed combinations that I might do. Will use different banks of the same midi instrument number, as you suggested.
Any advice on what I should do to sound the best? I recorded 96k 24-bit stereo pair for each, and took about .75-1.75 seconds of a note for each. I think I might try again and focus on always staying in tune and playing at a constant volume level so that the auto-loop feature can do its job the best...now that I know how that auto-loop feature works I notice that it does a better job when my sustain is as stable as possible. Otherwise the auto-loop unintentionally results a vibrato/tremolo.
Anybody know if there are other features of the sfz that I should take advantage of? I see there is a modulation configuration, but I don't know how that works or if it is useful.
(Note: I hear clicks sometimes in both the musescore wav output as well as during playback, interestingly the clicks don't always happen at the same places in the wav vs playback, and sometimes when I start playback at different location the clicks occur in different spots. So I can't figure out why the clicks occur, but would like to know if there is something I can do with this soundfont to fix them, although I think the clicks are something specific to the musescore synth.)
In reply to Thanks for your help. So far by ericfontainejazz
seems the playback is not looping the sample loop when on sustained notes. Maybe I missed something when I configured the soundfont...will try to figure out how to fix this.
Also, I've noticed that some of the pitches do not sound like they are the same volume level. I did use the polyphone noramlize function, but I think that is normalizing according to the maximum amplitude peak, and not the average power level of the sample. I think I'm going to try to redo and not even use the normalize function, but instead try to focus on playing all my notes at same volume, since the maximum peak amplitude is not representative of the average volume level.
Also, if the musescore synthesizer will attenuate/gain depending on dynamic levels, then when I want to make accented attack soundfont and pp & ff soundfonts, should I try to have the average power level of all the recorded samples to be roughly equal?
In reply to seems the playback is not by ericfontainejazz
Make sure you input for the loop to play (loop_playback in the instrument menu)
In reply to seems the playback is not by ericfontainejazz
You should normalise the samples in a sample editor such as Audacity before you start entering them into Polyphone.
IME Polyphone's auto-loop sucks.
You're better selecting a likely chunk of the waveform in Audacity and using the zero-crossing point finder to define the loop. Even then you will probably need to tweak it.
In reply to Thanks for your help. So far by ericfontainejazz
and while I'm asking questions, any thoughts on trying to make a vibrato version for each saxophone?
In reply to and while I'm asking by ericfontainejazz
Vibrato can be added within the soundfont, and also directly controlled by the Mod Wheel (Controler #1)
Personally I would use the soundfont facilities to add vibrato rather than using vibrato within the samples. - trying to tune them properly with vibrato is a bloody nightmare!
In reply to Vibrato can be added within by ChurchOrganist
ok...good advice!
In reply to Thanks for your help. So far by ericfontainejazz
It is pointless recording a saxophone in stereo.
If you think about it it is a monophonic sound source, and is easier to place within the stereo soundstage if it is recorded in mono.
Personally I always use 44.1kHz 16bit samples, as in any soundfont RAM consumption issues are important.
Recording 24bit 96kHz makes sense in a studio situation where the recording is going to a hard disk, but a soundfont is held in RAM, and even with today's huge RAM sizes it is still finite, and, if you are using instruments containing a sample per note - which you will want to do to retain quality, then you will soon start gobbling RAM space.
The new clarinet I'm working on for MuseScore is currently just under 8MB, and that contains just two sets of samples. Once I start adding velocity splits it will increase exponentially.
And if I add an Eflat clarinet as I have been requested to do, that will double its size.
In reply to It is pointless recording a by ChurchOrganist
I intend to reduce down to 44.1k 16bit mono, however I am recording at max setting I able to. I'm not actually able to tell the difference, but there are lots of people who claim that they can tell the difference and insist that they listen to stuff at the max bitrate, so I want to have a 96k 24bit stereo version for them.
Good to know you are working on a new clarinet sound. I will focus on my saxophones. I'd like to get mine into the next default musescore soundfont, so all your advice is welcome.
EDIT: approx where do you place your microphone(s). I have a pair of Oktava MK-012 and set them up at a stereo pair in my closet filled with clothes to help kill any stray sounds. Do you think I am better off recording with only one strategically placed microphone instead of stereo pair? Or maybe should I have one mic front center, maybe 50cm away from bell, and another off to the side where tone holes are, and combine to get a mix?
In reply to I intend to reduce down to by ericfontainejazz
I tend not to close mike. It is easier to capture the character of a wind instrument with one microphone placed in front of and above the instrument at a distance of around 3 feet.
But experiment and see what works best for you.