Why do SF2 soundfont presets sound so different to the samples?
I've been playing with a soundfont editor and have noticed that the final sound of the presets, as heard in MS, are noticeably different to the samples from which they are built. You can tell that the preset is based on the samples but it is nowhere near the same quality. I was honestly amazed at just how realistic the samples were. Is there something inherent in SF2s which causes this difference?
Comments
Maybe check your panning. Are Left samples panned fully Left and Right fully Right?
MuseScore by default adds reverb in the synthesizer settings and it can make the sample sound roomier. You can remove it, but it's best to keep some amount of reverb since most of the instruments come from different sources with different recording situations.
Thanks for the comments: they pointed me in the right direction. I checked the samples for correct panning and then the MS mixer settings for pan, reverb and chorus. All these were fine.
The volume in the mixer was 100 and the Master Gain was -40.00dB. I'm not sure what these are supposed to be but increasing them made the instrument in MS sound just like the samples so I reckon they must've been too low. The main difference is that the notes now ring properly whereas before they sounded truncated and a bit soft.
In reply to Thanks for the comments:… by yonah_ag
I do not mean the MIDI controller reverb that is found in the mixer. I mean the reverb in the synthesizer in the Master Effects tab. By default, MuseScore adds this effect which you can turn off by setting the Output to dry or by disabling the effect.
MuseScore also can cut off the sample if it's too long for the given note value which is why often times something written with a soundfont will sound unrealistic. The reason why the samples sound much more realistic is because it's a single performance with no context.
In reply to I do not mean the MIDI… by Quads are Awesome
Ah yes, these do make quite a difference. I hadn't used this tab of the synthesizer before. I've tweaked the reverb parameters, (only slightly), and this has improved things. I did try a totally dry mix but this sounded more artificial.
As you say, context also makes a difference so I think that this is probably as good as it's going to get - which is actually quite an improvement from where I started. The biggest difference came from using a dedicated guitar soundfont rather than using a GM soundfont.