CLI Flag for specifying bank offset

[derselbst]

Related discussion

#1473

Is your feature request related to a problem?

Loading multiple soundfonts with different bank offsets is currently not possible via the command line. The only option is to use a config file for fluidsynth's shell, which stops the MIDI player, loads the new Soundfont(s) with the desired offset and then restarts the player.

Describe the solution you'd like

Based on @mrbumpy409 's idea:

A new positional command-line flag --bank **n** / -b **n** would change the bank offset for any Sondfonts following that flag, e.g.:

fluidsynth -b 1 Altitude.sf2 Altitude.mid

...would load Altitude.sf2 into bank 1 and play Altitude.mid.

If you also wanted to specify GM/bank 0 SoundFont, you could do:

fluidsynth synthgs-sf2_04-compat.sf2 -b 1 Altitude.sf2 Altitude.mid

...which would load synthgs-sf2_04-compat.sf2 into the default bank 0, Altitude.sf2 into bank 1 and play Altitude.mid.

fluidsynth -b 1 Altitude.sf2 synthgs-sf2_04-compat.sf2 Altitude.mid

Would load both Soundfonts with a bankoffset of 1.

Describe alternatives you've considered

Another idea would be to simply use a prefix before the SoundFont path in "bank:SoundFont" format, e.g.:

fluidsynth 1:Altitude.sf2 Altitude.mid

If you need to use quotes for spaces in the filename, the quotes would start after the bank number prefix, e.g.:

fluidsynth 1:"Far From It.sf2" "Far From It.mid"

The drawback of this approach is, that it cannot be distinguished from a filename that really consists of a number and a colon at the beginning.

Additional context

Add any other context or screenshots about the feature request here.

[derselbst]

@mrbumpy409 FYI, I quickly smashed together a PR for the -b flag. Branch issue-1538 if you would like to check it out. E.g.

fluidsynth -v synthgs-sf2_04-compat.sf2 -b 1 Uplift-fixed.mid Uplift.sf2

(-v will tell you e.g. Now loading 'Uplift.sf2' with bank-offset=1)

With this change it will be no longer possible to specify the flags in an arbitrary order, e.g. you can no longer write

fluidsynth synthgs-sf2_04-compat.sf2 -R 0

Which I guess is fine, since the usage documentation has always been fluidsynth [options] [soundfonts] [midifiles]

[mrbumpy409]

This seems to work quite well. It looks like the order of the ".mid" and the bank 1 ".sf2" doesn't matter, correct? It is more logical to me to use the form:

fluidsynth synthgs-sf2_04-compat.sf2 -b 1 Uplift.sf2 Uplift-fixed.mid

...as it puts the bank offset command next to the ".sf2" file to be loaded into bank 1. This seems to work just fine for me.

The only issue I have so far is that the -b flag doesn't seem to work with the default SoundFont set in the FluidSynth config file. I like to have the GM bank specified in ~/.fluidsynth as follows:

set synth.default-soundfont "~/Documents/AWE32-midi-conversions/gm-soundfonts/synthgs-sf2_04-compat.sf2"

...but this seems to be ignored by FluidSynth when running:

fluidsynth -b 1 Uplift.sf2 Uplift-fixed.mid

It would be really useful to not have to manually specify the GM bank every time.

[derselbst]

It looks like the order of the ".mid" and the bank 1 ".sf2" doesn't matter, correct?

Correct. If a -b flag occurs, it simply says "now use this bank offset when loading the next soundfont(s)."

The only issue I have so far is that the -b flag doesn't seem to work with the default SoundFont set in the FluidSynth config file.

The scope of the -b flag is limited to the commandline options provided with the fluidsynth command. The synth.default-soundfont is a setting, that could come from anywhere (API, config file, default value compiled into fluidsynth, etc.). I don't see how both could be combined. Also note, fluidsynth only attempts to load the default-soundfont if no soundfont has been loaded. But since you have specified Uplift.sf2 via commandline, if would not load the GM bank. To load the GM bank with a bank offset you'd need to use load instead.

[mrbumpy409]

I'm sorry, I should have been more clear. I don't want to apply a bank offset to the GM bank; I want the GM bank to always load to bank 0 via the config file, and then "Uplift.sf2" is uploaded to bank 1 on top of that via the "fluidsynth" command. Loading "Uplift.sf2" with the bank offset instead cancels loading the default SoundFont from the config file, but I think the use of the offset flag for loading a SoundFont should instead be supplemental to synth.default-soundfont, rather than replacing it.

[derselbst]

Loading "Uplift.sf2" with the bank offset instead cancels loading the default SoundFont from the config file, but I think the use of the offset flag for loading a SoundFont should instead be supplemental to synth.default-soundfont, rather than replacing it.

I'm not sure if we understand each other correctly here. Even when not specifying the bank offset for Uplift.sf2, the default-soundfont would not be loaded. This behavior exists for a very long time now and it hasn't with introducing the bank offset.

I understand your use-case, but the only option I see is to put the load command into the config file. The config file will be processed before the MIDI player starts, so the player_stop / start commands we previously put in there should not be needed.

fluidsynth/src/fluidsynth.c

Lines 1085 to 1098 in 86197cb

	if(config_file != NULL && fluid_source(cmd_handler, config_file) < 0)
	{
	fprintf(stderr, "Failed to execute command configuration file '%s'\n", config_file);
	}
	/* start the player. Must be done after executing commands configuration file.
	This allows any existing player commands to be run prior the player is started.
	Example:
	player_tempo_bpm 60 # set a low tempo
	player_loop -1 # loop song forever
	*/
	if(player != NULL)
	{
	fluid_player_play(player);

[mrbumpy409]

Aah. Perhaps my described behavior seemed logical to me because that is how it works when playing AWE32 songs on Sound Blaster hardware, e.g., the GM bank is always loaded by default, and then you temporarily load the bank 1 SoundFont on top of that for playing the MIDI file.

I have been keeping a small selection of GM banks in the config file, and then I just uncomment the one I want to use, since dealing with long file paths is obnoxious:

#set synth.default-soundfont "/home/chris/Personal/Audio/Samples/SoundFonts/GeneralUser/GeneralUser-GS.sf2"
#set synth.default-soundfont "/home/chris/Personal/Documents/GitHub/mrbumpy409/AWE32-midi-conversions/gm-soundfonts/4gmgsmt-sf2_04-compat.sf2"
set synth.default-soundfont "/home/chris/Personal/Documents/GitHub/mrbumpy409/AWE32-midi-conversions/gm-soundfonts/synthgs-sf2_04-compat.sf2"

Same with the reverb and chorus settings, so I can switch between my usual preferences and my AWE32 MIDI preferences. However, I can make Linux scripts and aliases to accomplish what I want, so no biggie.

[derselbst]

I have been keeping a small selection of GM banks in the config file, and then I just uncomment the one I want to use, since dealing with long file paths is obnoxious:

Hm, unless I'm completely misreading the code, this shouldn't work. I need to have a closer look...

[mrbumpy409]

Here is my entire current ~/.fluidsynth file, if it's helpful:

#set audio.driver jack
set audio.driver pipewire
set audio.file.format float
set audio.file.type wav
set audio.jack.autoconnect true
set audio.period-size 64
set audio.periods 2
set audio.sample-format float
#set midi.portname "FluidSynth"
#set synth.default-soundfont "/home/chris/Personal/Audio/Samples/SoundFonts/GeneralUser/GeneralUser-GS.sf2"
#set synth.default-soundfont "/home/chris/Personal/Documents/GitHub/mrbumpy409/AWE32-midi-conversions/gm-soundfonts/4gmgsmt-sf2_04-compat.sf2"
set synth.default-soundfont "/home/chris/Personal/Documents/GitHub/mrbumpy409/AWE32-midi-conversions/gm-soundfonts/synthgs-sf2_04-compat.sf2"
#set synth.cpu-cores 2
set synth.device-id 16
#set synth.polyphony 512
set synth.sample-rate 48000
set synth.gain 0.5

#set synth.reverb.active 0
#set synth.chorus.active 0


# GeneralUser GS FX settings:
#set synth.reverb.damp 0.3
#set synth.reverb.level 0.7
#set synth.reverb.room-size 0.5
#set synth.reverb.width 0.8
#set synth.chorus.depth 3.6
#set synth.chorus.level 0.55
#set synth.chorus.nr 4
#set synth.chorus.speed 0.36

# AWE32 MIDI FX settings:
set synth.reverb.damp 0.3
set synth.reverb.level 0.6
set synth.reverb.room-size 0.65
set synth.reverb.width 0.8
set synth.chorus.depth 4.7
set synth.chorus.level 0.43
set synth.chorus.nr 3
set synth.chorus.speed 0.4

#interp 7

[derselbst]

I can confirm that putting set synth.default-soundfont /path/to/synthgs-sf2_04-compat.sf2 will not load the GM bank, if you any soundfont is explicitly specified. You must change it to load /path/to/synthgs-sf2_04-compat.sf2. You can verify it's being loaded by using the fonts command:

> fonts
ID  Name
 2  /path/to/synthgs-sf2_04-compat.sf2
 1  Uplift.sf2

The only difference is in which order the soundfonts are loaded onto fluidsynth's internal soundfont stack. When using the config file, the GM bank will be on the top of the stack. If you had specified it via commandline like so:

fluidsynth /path/to/synthgs-sf2_04-compat.sf2 Uplift.sf2

Uplift would be top on the stack:

> fonts
ID  Name
 2  Uplift.sf2
 1  /path/to/synthgs-sf2_04-compat.sf2

The order only matters if identical bank and presets exists for both soundfonts, but since Uplift is loaded with an offset, this doesn't matter for your use-case.

So in practice, you can simply use fluidsynth Uplift-fixed.mid -b 1 Uplift.sf2, just make sure to not use synth.default-soundfont in your config ;)

[mrbumpy409]

The only issue with this workaround is that any GM bank adhering to the Roland GS standard (such as GeneralUser-GS.sf2, synthgs-sf2_04-compat.sf2 and 4gmgsmt-sf2_04-compat.sf2) will have instruments on bank 1. Using the "load" method in the config file, these Roland GS bank 1 instruments (such as Solo French Horn, Synth Bass 101, etc.) will now overwrite any instruments in the song's custom SoundFont that happen to share the same preset number. So for playing AWE32 MIDI files, this is only an acceptable workaround if you are using a GM bank that does not contain such presets.

Note: Both of the Creative Labs banks only contain a subset of the GM sounds and only have the GS sound effects on bank 1. GeneralUser GS, on the other hand, contains all of the bank 1 instruments. Therefore the Creative Labs banks are less likely to overwrite presets of the song's custom SoundFont, but it is still possible.

[derselbst]

Ok, valid point. To address this, we would need to change the order when the config file is applied, to something like:

1. Apply the config file
2. Parse the commandline options.

Problem is that we only know which config file to load after we have parsed the commandline options (because a custom config may be provided via the -f flag)

Alternatively, we could change the way config files are loaded entirely. E.g. we could:

1. Always load the system config
2. Always load the user config
3. Parse the commandline options.
   a. If we see an -f option, apply that config file immediately before proceeding with the other commandline options.

But I can't tell if that breaks any other use-cases out there. If you have another idea, pls. let me know.

[mrbumpy409]

It would seem logical to me that any config file would be applied first—the "default" settings, if you will—and then command line options would override (or add to) what's in the config file. For example, if I have set synth.sample-rate 48000 in the config file, but pass -r 44100 on the command line, the command line option should override the setting in the config file. That's me coming at this from a non-programmer perspective and how I perceive config files vs command line options, but I don't know what others' expectations are.