I think the description on the original sfz v1 specs is a little ambiguous.
But the implementation of count in sfz.dll, sforzando and Cakewalk DropZone almost the same, except for the 0 and 1 behavior.
The region plays as many times as the count number

@kinwie should be OK now !

@paulfd Tested the new build for this. The count is fine now.

However, I encounter an issue for loop_count
The last portion of the sample after the loop is not played even I keep holding the key
Just in this simple example :

<region>
sample=snare_looped.wav
key=60
loop_count=3

After the loop, the snare will be cutoff. In Rapture and Sforzando, after the loop, it will play until the end of the sample.
Again, I know there is no detailed description for this opcode :))

Yes, in loop_sustain, the last portion of the sample after the loop, will be played after releasing the key and it controlled by ampeg_release

Okay so I revised all of this code to handle both the count, loop_count, and loop_mode with the differences between loop_sustain and loop_continuous.

I distinguished 3 cases:

• loop_continuous: only loop on the sample. Upon release, the sample will continue to loop. If loop_count is also set, the voice will release itself after loop_counts loops but still continue to loop during the release phase.
• loop_sustain: only loop on the sample until release, then play as if it were a normal sample. If loop_count is also set, the voice will release itself after loop_counts loops and play the part after the loop.
• one_shot: well you ignore loop_count in this case but count is taken into account. In all loop_ cases count is ignored.

If there are loop points in the sample file and no loop_mode in the SFZ file, we set loop_mode to loop_continuous and apply the above behavior. Maybe if loop_count is present setting the loop mode to loop_sustain would make more sense? It looks like it's what sforzando does, if loop_count is present the loop_mode is set to loop_sustain by default.

I created a test file with a kick, a looping hi-hat and a crash to test stuff if various players if you like.
loop_count.zip

I play your loop_count_test file in sforzando, DropZone and Rapture to compare to the current fix for loop_count.
Both sforzando and DropZone behave similar. I think in both, if loop_count is present loop_mode is set to loop_continuous, not loop_sustain.

From your test file, the loop_sustain is already fine.
Key d4 and f4 works similar like sforzando and DropZone.

Key e4 and g4 still not yet correct, imo, and I think the problem lies only at the loop_count behavior.
The behavior is, loop_count doesn't depend to the note-off message, unlike loop_sustain.

So,

• When a note is held as long as the sample or longer, it will play looped N times +1 and continue to the end portion of the sample (loop-end point to end point) and at full volume
• If the note released shorter than the sample and there is an ampeg_release, it will play looped N times +1 and continue to the end portion of the sample as long as value of ampeg_release (fade-out)

I've checked the "Fine tune the looping behavior" build. The loop_count is now working fine 👍