Where is the best place to learn about sampling?
-
- KVRist
- 136 posts since 31 Mar, 2002 from Scotland
-
Mental Audio Deviations Mental Audio Deviations https://www.kvraudio.com/forum/memberlist.php?mode=viewprofile&u=43630
- KVRist
- 180 posts since 7 Oct, 2004 from NL
You've found it. 
Just ask and look around here.
Just ask and look around here.Jaap
-
- KVRAF
- 13444 posts since 14 Nov, 2000 from Hannover / Germany
What would you need to know?
Real instrument sampling? Just sampling? Basic techniques?
Real instrument sampling? Just sampling? Basic techniques?
There are 3 kinds of people:
Those who can do maths and those who can't.
Those who can do maths and those who can't.
-
- KVRAF
- 6937 posts since 4 Jun, 2004 from Utrecht, Holland
Maybe a useful link: http://www.theprojectstudiohandbook.com/articles14.htm
-
- KVRAF
- 3002 posts since 24 Nov, 2003 from Heidelberg&Hamburg
Hi,
without owning Kontakt or Halion I liked to read on THAT website:
http://www.worldofsampling.com/gate.html?name=Content
There is a forum too, and some basic things are explained in tuts, which you don't only need for Kontakt or Halion.
without owning Kontakt or Halion I liked to read on THAT website:
http://www.worldofsampling.com/gate.html?name=Content
There is a forum too, and some basic things are explained in tuts, which you don't only need for Kontakt or Halion.
-
- KVRAF
- 13444 posts since 14 Nov, 2000 from Hannover / Germany
The SOS backissues (to be found in C00kie's link) usually are a good read.
Anyways, here's my three most important issues on sampling:
1) Creating/Recording the sample
- In case I sample natural sounds, do they have to be 24bit? If possible, I'd say yes, but for many sounds it might not be required. A sampled piano might sound grainy on its decay when only sampled in 24bit. Still about to sample a piano myself. Will see whenever and if I find some time.
- In case I sample VSTis (which I do a lot, for a whole number of reasons), Chainer is a godsend as it does all the recording and chopping job by itself.
- In case I sample something else (such as an external keyboard or Emagic's internal synths... which is what I'm doing right now), I set up some MIDI sequences, triggering the required notes.
- I allways try to decide WHICH part of a sound I actually need. A few things I may consider:
a) Sampling WITH LFOs (and other modulations) usually is no good idea unless you sample each and every note (chromatically that is), and even then the LFO tempo might later conflict with what I'm doing.
b) Sampling with filters sometimes is a great idea, but you'll have to know whether the filter is really unique enough to have it sampled. In case your sampler can reproduce it, let it do the job instead.
I usually prefer to open the filter a tad before sampling things because I can allways take it back a bit with some highcut inside the sampler.
In case some filter ADSR is applied, make sure it's necessary for the sound. I sometimes sample quick "zappy" filter ADSRs, later on I may then decide whether I really need the A (and D) part of it and cut things away (non-destructively, setting the start point later)
c) Amount of sampled keys. For simple sounds, you may be OK with them sampled in octaves. But be aware, in that case your sampler will have to do all the pitching work between the octaves, which isn't allways such a great idea, as it might result in aliasing artefacts. Personally, for simple sounds I usually go for a tritone (each C, each F# that is). For highly natural sounds you will have to raise that value. I usually decide between: Octaves, Tritones, major thirds, minor thirds, all white keys, chromatically.
d) Range of sampled keys. Obviously, this depends on the playability of the given instrument. I try to do it as broad as possible (for acoustic instruments anyways), In case you sample a lead synth it might not make too much sense going down all the way to C0 though.
e) Velocity of sampled keys. Tough to decide sometimes. The basic question being: Does the timbre change with higher/lower velocities or is the sound just getting a bit more brighter/duller? In case of the latter, your sampler might be able to reproduce that behaviour more or less fine (cutoff frequency triggered by velocity). In case the timbre is changing a lot (such as adding some extra brittle or buzzy sounds, like fret noises or the likes), you may want to sample more zones.
Actually, this is a VERY critical decision. Most drastic example: You only record two velocity zones, one clean version and another "buzzy" one. Fine. Now, when you map them up in your sampler, you may notice the switch between them while playing, instead of a seamingless transition (a problem with a lot of cheap patches and one of the reasons why sample libraries have been gotten so huge). You may think velocity x-fade between zones will fix this, but it doesn't work well quite sometimes as you will then be listening to two samples at the same time, which *might* cause some weird phasing.
So, in case you're sampling multiple velocities I recommend sampling as much as possible, in oder to keep velocity x-fades small or even avoid them at all. Way more work though.
f) Recording levels. Quite a difficult thing IMO. Obviously, the loudest notes should be recorded at maximum level, but what about the lower velocity zones? Tough question IMO - there seem to be people raising record level, but IMO that's not such a good idea. I'd setup recording levels for the loudest hit and then get away with it. Will make your samples sound more naturally.
2) Sample treatment
- You will have to have something cutting up your recordings properly. Not a big deal. Just, when saving the individual note samples (which I would do, I wouldn't cut them up inside the zone editor of the sampler), make sure to follow certain naming conventions, either made up by yourself or following some standards.
My naming convention almost allways goes like this:
"XYZ01_V127_C0.wav"
XYZ = Name of sound, e.g. "BrightPiano"
01 = Number of variation.
V127 = Velocity used. For natural samples those numbers can only be approximate ones. They'll help you to map up things more quickly though.
C0 = MIDI note name. Note: International MIDI conventions are like C0, F#1, etc. No flats (can be confused with a letter B), no "-". some samplers support automapping, following note namings, therefor I allways add them properly.
I use underscores because I found spaces not to work on some occasions.
- Time for the most critical decision now: Loop those samples or don't? I actually add looppoints to ALL my samples, if possible at least (note: those looppoints are embedded inside the file header and samplers will read them out fine). You will need an editor allowing to set looppoints, preferably with an automated search option. WaveLab is what I'm using but I heard pretty good things about Zero-X Seamless Looper which is WAY cheaper.
You may now ask: Why would I loop a sample? For certain reasons: Imagine you'd just have a simple square wave patch or so. It'd look all the same all over 5 minutes (in case you'd need it for a long sustained pad or whatever), so why waste HDD space and RAM? Instead you'd just loop it and take, say, 1 or 2 samples only. That way a simple squarewave patch would only take up a few KB.
Looping is defenitely getting more difficult with natural sounds. You will have to decide whether the non-looped sound is sufficient (i.e. ringing long enough anyways) In case it isn't, in case it has some relatively quick decay, you may not be able to find a proper seamless looppoint because you may be able to notice the level differences from loop-end to loop-start. Too bad.
The solution here might be compression. But careful!!! This will destructively alter your sample, so you better "save as" or work on a copy.
Try with a very long attack value, so the natural attack isn't affected (this is THE most important portion in all natural samples), use a relatively high compression ratio and release time. The optimal thing should be a sample which looks like the natural attack phase going into a very even looking decay.
In addition to that you may add some crossfading. This will make transitions from loop-end to loop-start WAY easier. For natural sounds I'd try with a relatively short value. X-fades usually can be done in your sampler as well, those are non destructive, so you may try those first (even if your loop points sound a bit like a hickup for now).
For sampled synth-ish, somewhat modulating sounds I often use a really large x-fade value. I may not need the dacay and attack phases of those at all.
Which reminds me of: To save space (HDD and RAM), I often cut everything right from the looppoints away, because I may rebuild the decay with my sampler's ADSR anyways. For sounds with un-characteristic attacks I may cut away everything left from my loop points as well.
Note: For truly natural sounds looping, compressing and x-fading aren't an option anymore. HDD space is rather cheap and most samplers support disk streaming these days too.
But in case you plan to use your samples for something a bit more experimental (such as making a pad out of your piano), looping might be useful for them as well.
3) Mapping that shit up in your sampler.
Not much to say, all depending too much on your sampler.
Personally I'm using the EXS for all my basic mappings because it's got that nice automap feature I mentioned above.
No matter which sampler you're using, make sure to either use separate groups for the individual velocity zones/samples. I sometimes even build temporary patches containing JUST ONE velocity zone and paste them together later on. Depends on what soret of functionality your sampler supports (in Kontakt it's rather easy to only have one group shown), but no matter what (and how) you do, separating velocity zones/groups (or complete "sound zones/groups", should you plan to layer sounds) will make maintainance WAY more easy!
Last thing would be to carefully rebuild some naturally sounding amp ADSR (in case that's what you're after). As mentioned above, some very old trick to make things sound more natural might be to control some soft filter cutoff by velocity. Most natural instruments tend (or seem) to sound brighter on higher volumes and controlling a filter CO with velocity is a pretty much usual way to emulate this (even not so natural patches might sound more organic that way).
Phew, that's got to be my sampling 101 or so.
Anyways, here's my three most important issues on sampling:
1) Creating/Recording the sample
- In case I sample natural sounds, do they have to be 24bit? If possible, I'd say yes, but for many sounds it might not be required. A sampled piano might sound grainy on its decay when only sampled in 24bit. Still about to sample a piano myself. Will see whenever and if I find some time.
- In case I sample VSTis (which I do a lot, for a whole number of reasons), Chainer is a godsend as it does all the recording and chopping job by itself.
- In case I sample something else (such as an external keyboard or Emagic's internal synths... which is what I'm doing right now), I set up some MIDI sequences, triggering the required notes.
- I allways try to decide WHICH part of a sound I actually need. A few things I may consider:
a) Sampling WITH LFOs (and other modulations) usually is no good idea unless you sample each and every note (chromatically that is), and even then the LFO tempo might later conflict with what I'm doing.
b) Sampling with filters sometimes is a great idea, but you'll have to know whether the filter is really unique enough to have it sampled. In case your sampler can reproduce it, let it do the job instead.
I usually prefer to open the filter a tad before sampling things because I can allways take it back a bit with some highcut inside the sampler.
In case some filter ADSR is applied, make sure it's necessary for the sound. I sometimes sample quick "zappy" filter ADSRs, later on I may then decide whether I really need the A (and D) part of it and cut things away (non-destructively, setting the start point later)
c) Amount of sampled keys. For simple sounds, you may be OK with them sampled in octaves. But be aware, in that case your sampler will have to do all the pitching work between the octaves, which isn't allways such a great idea, as it might result in aliasing artefacts. Personally, for simple sounds I usually go for a tritone (each C, each F# that is). For highly natural sounds you will have to raise that value. I usually decide between: Octaves, Tritones, major thirds, minor thirds, all white keys, chromatically.
d) Range of sampled keys. Obviously, this depends on the playability of the given instrument. I try to do it as broad as possible (for acoustic instruments anyways), In case you sample a lead synth it might not make too much sense going down all the way to C0 though.
e) Velocity of sampled keys. Tough to decide sometimes. The basic question being: Does the timbre change with higher/lower velocities or is the sound just getting a bit more brighter/duller? In case of the latter, your sampler might be able to reproduce that behaviour more or less fine (cutoff frequency triggered by velocity). In case the timbre is changing a lot (such as adding some extra brittle or buzzy sounds, like fret noises or the likes), you may want to sample more zones.
Actually, this is a VERY critical decision. Most drastic example: You only record two velocity zones, one clean version and another "buzzy" one. Fine. Now, when you map them up in your sampler, you may notice the switch between them while playing, instead of a seamingless transition (a problem with a lot of cheap patches and one of the reasons why sample libraries have been gotten so huge). You may think velocity x-fade between zones will fix this, but it doesn't work well quite sometimes as you will then be listening to two samples at the same time, which *might* cause some weird phasing.
So, in case you're sampling multiple velocities I recommend sampling as much as possible, in oder to keep velocity x-fades small or even avoid them at all. Way more work though.
f) Recording levels. Quite a difficult thing IMO. Obviously, the loudest notes should be recorded at maximum level, but what about the lower velocity zones? Tough question IMO - there seem to be people raising record level, but IMO that's not such a good idea. I'd setup recording levels for the loudest hit and then get away with it. Will make your samples sound more naturally.
2) Sample treatment
- You will have to have something cutting up your recordings properly. Not a big deal. Just, when saving the individual note samples (which I would do, I wouldn't cut them up inside the zone editor of the sampler), make sure to follow certain naming conventions, either made up by yourself or following some standards.
My naming convention almost allways goes like this:
"XYZ01_V127_C0.wav"
XYZ = Name of sound, e.g. "BrightPiano"
01 = Number of variation.
V127 = Velocity used. For natural samples those numbers can only be approximate ones. They'll help you to map up things more quickly though.
C0 = MIDI note name. Note: International MIDI conventions are like C0, F#1, etc. No flats (can be confused with a letter B), no "-". some samplers support automapping, following note namings, therefor I allways add them properly.
I use underscores because I found spaces not to work on some occasions.
- Time for the most critical decision now: Loop those samples or don't? I actually add looppoints to ALL my samples, if possible at least (note: those looppoints are embedded inside the file header and samplers will read them out fine). You will need an editor allowing to set looppoints, preferably with an automated search option. WaveLab is what I'm using but I heard pretty good things about Zero-X Seamless Looper which is WAY cheaper.
You may now ask: Why would I loop a sample? For certain reasons: Imagine you'd just have a simple square wave patch or so. It'd look all the same all over 5 minutes (in case you'd need it for a long sustained pad or whatever), so why waste HDD space and RAM? Instead you'd just loop it and take, say, 1 or 2 samples only. That way a simple squarewave patch would only take up a few KB.
Looping is defenitely getting more difficult with natural sounds. You will have to decide whether the non-looped sound is sufficient (i.e. ringing long enough anyways) In case it isn't, in case it has some relatively quick decay, you may not be able to find a proper seamless looppoint because you may be able to notice the level differences from loop-end to loop-start. Too bad.
The solution here might be compression. But careful!!! This will destructively alter your sample, so you better "save as" or work on a copy.
Try with a very long attack value, so the natural attack isn't affected (this is THE most important portion in all natural samples), use a relatively high compression ratio and release time. The optimal thing should be a sample which looks like the natural attack phase going into a very even looking decay.
In addition to that you may add some crossfading. This will make transitions from loop-end to loop-start WAY easier. For natural sounds I'd try with a relatively short value. X-fades usually can be done in your sampler as well, those are non destructive, so you may try those first (even if your loop points sound a bit like a hickup for now).
For sampled synth-ish, somewhat modulating sounds I often use a really large x-fade value. I may not need the dacay and attack phases of those at all.
Which reminds me of: To save space (HDD and RAM), I often cut everything right from the looppoints away, because I may rebuild the decay with my sampler's ADSR anyways. For sounds with un-characteristic attacks I may cut away everything left from my loop points as well.
Note: For truly natural sounds looping, compressing and x-fading aren't an option anymore. HDD space is rather cheap and most samplers support disk streaming these days too.
But in case you plan to use your samples for something a bit more experimental (such as making a pad out of your piano), looping might be useful for them as well.
3) Mapping that shit up in your sampler.
Not much to say, all depending too much on your sampler.
Personally I'm using the EXS for all my basic mappings because it's got that nice automap feature I mentioned above.
No matter which sampler you're using, make sure to either use separate groups for the individual velocity zones/samples. I sometimes even build temporary patches containing JUST ONE velocity zone and paste them together later on. Depends on what soret of functionality your sampler supports (in Kontakt it's rather easy to only have one group shown), but no matter what (and how) you do, separating velocity zones/groups (or complete "sound zones/groups", should you plan to layer sounds) will make maintainance WAY more easy!
Last thing would be to carefully rebuild some naturally sounding amp ADSR (in case that's what you're after). As mentioned above, some very old trick to make things sound more natural might be to control some soft filter cutoff by velocity. Most natural instruments tend (or seem) to sound brighter on higher volumes and controlling a filter CO with velocity is a pretty much usual way to emulate this (even not so natural patches might sound more organic that way).
Phew, that's got to be my sampling 101 or so.
There are 3 kinds of people:
Those who can do maths and those who can't.
Those who can do maths and those who can't.
-
- Hun #3
- 4265 posts since 25 Mar, 2002 from A quaint little village just south of Hamburg, Germany
Like the others said...MacNastie wrote:Looking for links, best software, best ways to use tips tricks etc
just post any question in everything else and see the answers materialize
-
- KVRist
- Topic Starter
- 136 posts since 31 Mar, 2002 from Scotland
You talk to me like I have never been here before....mainly I am being lazy and want to know how to use a sampler rather than actually recording samples.....how do you put a sample into the sampler, are disks set up in such a way that it loads all samples for each not or does the sampler create each note from one sample?
More questions to follow i'm sure.... i have tried before to use samplers but it was years ago.
More questions to follow i'm sure.... i have tried before to use samplers but it was years ago.
-
- KVRian
- 534 posts since 18 Mar, 2002 from france
-
- KVRAF
- 6937 posts since 4 Jun, 2004 from Utrecht, Holland
That depends on which sampler you use (hardware, software, brand & type) Record from line in, or feed WAV/AIFF from disk.MacNastie wrote:how do you put a sample into the sampler
In general it will load all samples nescessary for a patch into memory (unless its gigabytes instead of megabytes and disk streaming is used)MacNastie wrote:are disks set up in such a way that it loads all samples for each not or does the sampler create each note from one sample?
Also this depends on how the patch is set up. In a simple patch there is one sample in memory which is resampled on-the-fly when played for a different pitch. There can be different samples used for different keyboard ranges so it doesn't transpose more than a couple of semitones. This increases size of the patch drastically but results in a natural sound.
More questions?? Just have a play with it...