SF32 Mode
It works when I load the SF2 file as PR 32. What does that mean?
I found a manual at
Therein it says:
SF32 mode
In SF32 mode, the sfz player will load the whole SoundFont Bank file into memory (that is,
all the waveforms for all of the programs in the SoundFont Bank will be loaded). The waves
will be converted to 32-bit during loading.
This option optimizes performance and flexibility so programs can be changed while voices
are playing without interruption. However, this option uses a lot of memory—approximately
two-and-a-half times the size of the SoundFont Bank.
If there is insufficient memory, then the sfz player will switch to SF16 mode (see the fol-
lowing section), which will reduce the amount of memory required by about 50%. However,
this also increases the CPU overhead.
PR32 Mode
In PR32 mode, only the samples required by the current program (preset) are loaded and
converted to 32-bit. This mode only requires memory for the samples that will be used, hence
the RAM requirement may be reduced considerably (depending on the size of the SoundFont
Bank).
When running in PR32 mode, program changes will lead to all voices being stopped imme-
diately before the next set of samples is loaded. This might lead to unwanted artifacts during
playback.