The most future proof format for sampled instruments is SFZ
- KVRAF
- 20768 posts since 22 Nov, 2000 from Southern California
- KVRAF
- Topic Starter
- 7111 posts since 19 Apr, 2002 from Utah
To further the conversation, this would be an example of a very, very basic multisampled instrument that I might create with SFZ would be the ideal for future proofing of my personal sampled instrument:
* Record at least one set of multisamples samples dry so that effects can be added/changed later, post sampler
* Mono (easier to loop, uses less space, and the stereo effect can be added/changed later with a plugin, post sampler
* 24bit/48khz
* 15 sec. note length per sample
* Normalized
* 7 octaves
* 1 Velocity
* Sample C, E, and G# per octave
* 21 samples total
* Use a naming convention. For example: <name>_<Note/Octave>_<xxxBpm>.WAV like this: spacesynth_C3_120Bpm.WAV
One could sample as deeply as desired, of course, but for the casual user, this might be an example of a typical multisampled instrument that could very easily be loaded into any sampler in the world from a SFZ instrument with very little tweaking needed on the new sampler being used.
* Record at least one set of multisamples samples dry so that effects can be added/changed later, post sampler
* Mono (easier to loop, uses less space, and the stereo effect can be added/changed later with a plugin, post sampler
* 24bit/48khz
* 15 sec. note length per sample
* Normalized
* 7 octaves
* 1 Velocity
* Sample C, E, and G# per octave
* 21 samples total
* Use a naming convention. For example: <name>_<Note/Octave>_<xxxBpm>.WAV like this: spacesynth_C3_120Bpm.WAV
One could sample as deeply as desired, of course, but for the casual user, this might be an example of a typical multisampled instrument that could very easily be loaded into any sampler in the world from a SFZ instrument with very little tweaking needed on the new sampler being used.
Vendor‑Dependent Copy Protection: Customers lose. Pirates win.
(Also: I'm Accused of lying about Linux—it boots, runs my pro audio workflow, stays stable, updates--though yearly dismissed as “niche”. Yet I'm the deluded one.)
(Also: I'm Accused of lying about Linux—it boots, runs my pro audio workflow, stays stable, updates--though yearly dismissed as “niche”. Yet I'm the deluded one.)
- KVRAF
- 20768 posts since 22 Nov, 2000 from Southern California
I do A-1, F#0, and A0, then 4 notes per octave from C1 to C5 (C, D#, F#, and A). I stop at C5 because anything higher is unnecessary and often unpleasant sounding. I normalize at -12 but -6 is usually fine. Looping stereo files isn't a problem, especially when using back-and-forth looping.
-
- KVRAF
- 3358 posts since 19 Mar, 2008 from germany
So your examples "Legato-Samples" and "Round Robin"Burillo wrote: Sat Jun 17, 2023 1:36 pm Is sfz anywhere as capable as something like Kontakt is? Like, can it do legato samples, round robin, and other complex scripting? I mean, the base assumption behind this thread is that loop points and wav samples is all you need from a sampler, but that's not true for like 90% of instrument sample libraries out there.
are easily realizable with sfz. But also something like
"Auto-Mute" and various "Filter controls via velocity or
controller --> velocity, LFO, cutoff, filter, amplifier, sample-
starts" etc. can be extensively implemented or are actually
standard.
What doesn't work compared to Kontakt is a variable definition
(see contribution by DSmolken). You could of course do a lot
with it - everyone has to decide for themselves whether they
really need it.
What's important to me personally is that the sample format
is non-proprietary. Anyone who was dependent on a
proprietary format like us - and once spent days looking for
the activation key and is desperate because otherwise they
would not be able to access certain samples, will understand
this. And from then on, swear to yourself:
Only open sample formats!
Last edited by enroe on Mon Jun 19, 2023 1:57 am, edited 1 time in total.
free mp3s + info: andy-enroe.de songs + weird stuff: enroe.de
- KVRAF
- Topic Starter
- 7111 posts since 19 Apr, 2002 from Utah
What I posted was just an example of a possible mapping scenario that could be used for a future proof sfz instrument. There’s nothing wrong with the way you do it. In fact, it may be a better example than mine.Uncle E wrote: Sun Jun 18, 2023 4:32 am I do A-1, F#0, and A0, then 4 notes per octave from C1 to C5 (C, D#, F#, and A). I stop at C5 because anything higher is unnecessary and often unpleasant sounding. I normalize at -12 but -6 is usually fine. Looping stereo files isn't a problem, especially when using back-and-forth looping.
Vendor‑Dependent Copy Protection: Customers lose. Pirates win.
(Also: I'm Accused of lying about Linux—it boots, runs my pro audio workflow, stays stable, updates--though yearly dismissed as “niche”. Yet I'm the deluded one.)
(Also: I'm Accused of lying about Linux—it boots, runs my pro audio workflow, stays stable, updates--though yearly dismissed as “niche”. Yet I'm the deluded one.)
- KVRAF
- 7412 posts since 8 Feb, 2003 from London, UK
Not exactly but with the sw_ opcodes, you do get some state maintained. It's not a procedural or functional scripting language, I'll give you that but there is a state machine and you get to control what it's doing, so you're feeding that a script, essentially.DSmolken wrote: Sat Jun 17, 2023 7:55 pm Pedantically: it's not a scripting language as such, it's an event filter configuration.
It is possible to do multi-mic samples, true or fake legato, and a lot of other stuff. It's not possible (at least in current players and as far as I know) to have variables that have persistence between notes or free-running LFOs.
- KVRAF
- 20768 posts since 22 Nov, 2000 from Southern California
Sure, I just posted it in case it might give you ideas, especially regarding notes above C5.audiojunkie wrote: Sun Jun 18, 2023 7:56 am What I posted was just an example of a possible mapping scenario that could be used for a future proof sfz instrument.
- KVRAF
- Topic Starter
- 7111 posts since 19 Apr, 2002 from Utah
Actually, I’d love to hear your thoughts!Uncle E wrote: Sun Jun 18, 2023 9:14 pmSure, I just posted it in case it might give you ideas, especially regarding notes above C5.audiojunkie wrote: Sun Jun 18, 2023 7:56 am What I posted was just an example of a possible mapping scenario that could be used for a future proof sfz instrument.
Vendor‑Dependent Copy Protection: Customers lose. Pirates win.
(Also: I'm Accused of lying about Linux—it boots, runs my pro audio workflow, stays stable, updates--though yearly dismissed as “niche”. Yet I'm the deluded one.)
(Also: I'm Accused of lying about Linux—it boots, runs my pro audio workflow, stays stable, updates--though yearly dismissed as “niche”. Yet I'm the deluded one.)
- KVRAF
- Topic Starter
- 7111 posts since 19 Apr, 2002 from Utah
Of course!
SFZ supports most of the popular formats. 
Vendor‑Dependent Copy Protection: Customers lose. Pirates win.
(Also: I'm Accused of lying about Linux—it boots, runs my pro audio workflow, stays stable, updates--though yearly dismissed as “niche”. Yet I'm the deluded one.)
(Also: I'm Accused of lying about Linux—it boots, runs my pro audio workflow, stays stable, updates--though yearly dismissed as “niche”. Yet I'm the deluded one.)
- KVRAF
- Topic Starter
- 7111 posts since 19 Apr, 2002 from Utah
I'm not sure if you saw my previous post. I would love to hear your thoughts on notes above C5.Uncle E wrote: Sun Jun 18, 2023 9:14 pmSure, I just posted it in case it might give you ideas, especially regarding notes above C5.audiojunkie wrote: Sun Jun 18, 2023 7:56 am What I posted was just an example of a possible mapping scenario that could be used for a future proof sfz instrument.
Vendor‑Dependent Copy Protection: Customers lose. Pirates win.
(Also: I'm Accused of lying about Linux—it boots, runs my pro audio workflow, stays stable, updates--though yearly dismissed as “niche”. Yet I'm the deluded one.)
(Also: I'm Accused of lying about Linux—it boots, runs my pro audio workflow, stays stable, updates--though yearly dismissed as “niche”. Yet I'm the deluded one.)
- KVRAF
- 20768 posts since 22 Nov, 2000 from Southern California
Oh, sorry, I didn't realize you wanted me to respond. My experience has been that notes above C5 pitch just fine and don't need to be multi-sampled. But I should have mentioned that piano, violin, flute, or other instruments where very high notes are commonly used are a different story.audiojunkie wrote: Fri Jun 23, 2023 6:49 pm I'm not sure if you saw my previous post. I would love to hear your thoughts on notes above C5.I apologize if I seemed dismissive of your original post. It wasn't my intention. I'm always open to learn new tricks, and I'm really curious now why you don't use notes above C5.
![]()
- KVRAF
- Topic Starter
- 7111 posts since 19 Apr, 2002 from Utah
That makes sense. For example, there is no reason to capture that high for a bass instrument. 
Vendor‑Dependent Copy Protection: Customers lose. Pirates win.
(Also: I'm Accused of lying about Linux—it boots, runs my pro audio workflow, stays stable, updates--though yearly dismissed as “niche”. Yet I'm the deluded one.)
(Also: I'm Accused of lying about Linux—it boots, runs my pro audio workflow, stays stable, updates--though yearly dismissed as “niche”. Yet I'm the deluded one.)
- KVRian
- 792 posts since 9 Feb, 2019
Notes above C5 are not as different anymore as around C3, but synths sound very different at higher pitch. Analog synths often sound interestingly metallic and you can generate interesting FX or percussion sounds.
When it´s about synths we should also consider suboscillators and layers and you cam also transpose them if the sound is used as a layer itself.
I usuallly sample C-0 - C8 and eventually leave away what I think is unnecessary.
You have no Bass at C5 anymore, but a sometimes maybe even more interesting sound to work with.
When it´s about synths we should also consider suboscillators and layers and you cam also transpose them if the sound is used as a layer itself.
I usuallly sample C-0 - C8 and eventually leave away what I think is unnecessary.
You have no Bass at C5 anymore, but a sometimes maybe even more interesting sound to work with.
- KVRAF
- Topic Starter
- 7111 posts since 19 Apr, 2002 from Utah
I believe a lot of that metallic sound is aliasing. As I understand it, it has a lot to do with the sample rate used. Sampling at a much higher rate allows for the alias noise to be filtered off above the hearing threshold. I know it's not so important for recording anything that is not transposed--for example, recording normal audio at 24bit/48khz will always likely be sufficient. But in the case of transposing audio, transposing a 44.1khz sample or lower would put the aliasing noise right across the audible frequencies. The more I read about this stuff, the more it makes me think differently about how I would sample things. For example, I'm thinking more now about doing things this way for chromatic instruments, to keep things easy, the quality high, and the work minimal:
* Use a generic SFZ map template that covers 4 notes per octave, with 5 octaves--using octaves 2 through 5.
* Record at least one set of multisamples dry so that effects can be added/changed later, with plugins
* Record Mono (Easier to loop, and smaller--also, the stereo effect can be added/changed later using different means)
* Record 24bit/192Khz to maintain audio quality during interpolation process used during transposition, and also to keep aliasing noise to a much higher frequency outside of the audible range.
* Record 15 seconds note length per sample (This allows for enough space to get a good loop, and can be cut/trimmed later).
* Balance the volume levels between the various multisamples as needed, and then normalize each sample
* Use 5 octaves (the 2nd to the 5th octaves, and stretch the first sample zone (C2) to cover C1 and below, and stretch the last sample zone (C5) to cover C7 and above)
* Determine the primary instrument emphasis (bass instrument, mid-range, treble instrument). If it is a bass instrument, start mapping everything transposed to the start of the C2 octave. For example, start mapping the C1 sample at C2 so that it fits in the 2nd to 5th octaves map. The same with a treble instrument, except the map starts with mapping the C7 sample on the C5 key. This way, there are 5 good octaves of sounds from a generic SFZ template.
* Use just 1 velocity layer, and have the volume controlled through velocity strength
* Sample C, D#, F#, and A per octave, for 5 octaves (20 samples total)
* Use a naming convention. For example <name>_<Descriptor>_<Note/Octave>_<xxxBpm>.WAV like this: Spacesynth_Dry_C3.WAV or Drumloop_Ambient_110Bpm.WAV
I'll likely refine this further as I learn more tricks, but the general idea is to record, name, and loop the multisamples so that they can easily be imported into nearly any sampler.
* Use a generic SFZ map template that covers 4 notes per octave, with 5 octaves--using octaves 2 through 5.
* Record at least one set of multisamples dry so that effects can be added/changed later, with plugins
* Record Mono (Easier to loop, and smaller--also, the stereo effect can be added/changed later using different means)
* Record 24bit/192Khz to maintain audio quality during interpolation process used during transposition, and also to keep aliasing noise to a much higher frequency outside of the audible range.
* Record 15 seconds note length per sample (This allows for enough space to get a good loop, and can be cut/trimmed later).
* Balance the volume levels between the various multisamples as needed, and then normalize each sample
* Use 5 octaves (the 2nd to the 5th octaves, and stretch the first sample zone (C2) to cover C1 and below, and stretch the last sample zone (C5) to cover C7 and above)
* Determine the primary instrument emphasis (bass instrument, mid-range, treble instrument). If it is a bass instrument, start mapping everything transposed to the start of the C2 octave. For example, start mapping the C1 sample at C2 so that it fits in the 2nd to 5th octaves map. The same with a treble instrument, except the map starts with mapping the C7 sample on the C5 key. This way, there are 5 good octaves of sounds from a generic SFZ template.
* Use just 1 velocity layer, and have the volume controlled through velocity strength
* Sample C, D#, F#, and A per octave, for 5 octaves (20 samples total)
* Use a naming convention. For example <name>_<Descriptor>_<Note/Octave>_<xxxBpm>.WAV like this: Spacesynth_Dry_C3.WAV or Drumloop_Ambient_110Bpm.WAV
I'll likely refine this further as I learn more tricks, but the general idea is to record, name, and loop the multisamples so that they can easily be imported into nearly any sampler.
Vendor‑Dependent Copy Protection: Customers lose. Pirates win.
(Also: I'm Accused of lying about Linux—it boots, runs my pro audio workflow, stays stable, updates--though yearly dismissed as “niche”. Yet I'm the deluded one.)
(Also: I'm Accused of lying about Linux—it boots, runs my pro audio workflow, stays stable, updates--though yearly dismissed as “niche”. Yet I'm the deluded one.)
