hi,
dunno if this is a current issue anymore or if it's really really the case at all in your software:
using the sf2 part of your software (the renoise tool "sf2xrni" uses it), the first sample (some left C-3) has 2 to incorrect additional values, and probably all samples have it too but just go into the zero-padding of the previous one
at least that's what the output of sf2xrni looked like
when i made my own sf2 to xrni code, i did at first not read the 32 bit integer coming after the "smpl" chunk id, and i got the same result
i changed my code to read a 32 bit chunk size after "smpl" and it a) seems to be the sample data size and doesn't break parsing b) the first sample is now correct
dunno if this is a current issue anymore or if it's really really the case at all in your software:
using the sf2 part of your software (the renoise tool "sf2xrni" uses it), the first sample (some left C-3) has 2 to incorrect additional values, and probably all samples have it too but just go into the zero-padding of the previous one
at least that's what the output of sf2xrni looked like
when i made my own sf2 to xrni code, i did at first not read the 32 bit integer coming after the "smpl" chunk id, and i got the same result
i changed my code to read a 32 bit chunk size after "smpl" and it a) seems to be the sample data size and doesn't break parsing b) the first sample is now correct