Good enough settings to make sample library ?

[hoanglongdummy]

I plan to make sample library, so would the settings like below be enough to use in almost music and sound production ?

Sampling quality 128 point sync
44100 Hz
16 bit depth
stereo
record only from C4 to C7 (Based on Fl Studio key range C0 - C10, and to save some storage)
file format .wav

Record using DirectWave from Fl Studio

Record length may vary on samples and it have an effect on storage.
What do you guys think ? I prioritize good samples and low storage.

[Obsolete462444]

Some factors that should be considered as well:

- how many velocity layers?
- are you going to sample each, every 2nd, 3rd ... semitone or only one note for each octave?
- are you going to include round robins?

And if you sample an acoustic instrument, you should definitely sample the full length of the tone.

Here is one approach: https://www.pianobook.co.uk/how-to-sample/

[DSmolken]

Basically, yes, 44.1 kHz stereo and 16 bits is a lot for a lot of applications. If you're doing things that will get a lot of compression or distortion applied by the user, you should probably also do 24 bits instead of 16. And additional mic positions beyond stereo are useful for acoustic drum kits, orchestral instruments and other things where the room is an important part of the overall sound. Recording length should vary... you don't need 40 seconds for a snare drum hit, though I've seen samples like that. On the other hand, you might need 40 seconds for a deep boom, or a ride cymbal.

And as mentioned in the first reply, things like round robins and dynamic layers, plus articulations and possibly legato are also very very important considerations for acoustic instruments.

[GRUMP]

24/48 is Standard atM and you can convert it 16/44 later on or offer different Qualities.

You should not worry about File Sizes. The Recording should better focus on Quality and Flexibility. You can save Space later - but me personally - I suspect that nobody will care about that anymore in some Years.

The Rest depends. I'd recommend to decide in the single Case, regarding the concrete Sound in Production. You should Always focus on the Sound and how you can record, reproduce, handle, ... it the best Way. There are Always various Options.

And Samplers also offer the Possibility to go further than the given Sound. They have a creative Potential. I strongly recommend to consider that from the Beginning and make the Product a unique Sampler INSTRUMENT (and not just a sampled Synth Sound).

Long Story ...

Export don't record and concider Mono!

[lfm]

If using, as said, 24 bit you do not have to worry so much to keep peaks close to 0 dBFs, you can without loss raise that later and still have way more distance to noisefloor. As 24 bit record with peaks -12 dBFs and no worries, but if doing that 16 bit you loose a lot of signal noise ratio.

Remember that most A/D converters are 20 bits in reality.

Spitfire Audio focusing on video score deliver samples in 48k which is standard for video, so no cpu needs to be used doing on the fly SRC while using samples. So one thing to consider.

[hoanglongdummy]

Thanks, all of these are good advices

Unfortunately, DirectWave don't let me easily record the samples in 24 bit depth. I may have to settle with 32 bit float then, oh my poor computer and hard drive.

Damn, make this sample library low storage is not an option. Have to make this sample library worthwhile then.

[BertKoor]

You asked for good enough? 16bit is good enough and smaller than 24bit but not that much worse.
Record as 32bit float then (technically not that different from 24bit) and convert to 16bit at the last possible moment.

Don't record mono sounds as stereo!

[Obsolete462444]

Yes, use 32 bit then. Size should not be an issue these days. Once all the samples are processed / edited you can batch convert to 16 bit using this great free tool (high quality sample rate conversion):

https://www.voxengo.com/product/r8brain/

[JeffLearman]

Definitely record in 24 or 32 bit format, and postprocess to put in the format you want. Once you're done, if you use normalized velocity layers (which I believe is best in most cases), you can convert to 16 bit and the sound quality is excellent. If you decide not to normalize each velocity layer, then you'll either need a LOT of velocity layers, or you'll need to understand how to program the dynamics in a nontrivial way.

Always sample the range of the original instrument, if there is an "original instrument." Don't limit yourself to C4-C7 unless that's the normal range for whatever you're sampling.

What are you sampling, btw? What format do you plan to use (sfz, kontakt, etc.)?

You'll end up with lots of files. I find it really helps to use SoX audio processor https://sourceforge.net/projects/sox/ which is one command-line program that does pretty much anything you can think of in the way of audio processing (converting formats, normalizing, resampling, filtering, etc.) It's probably the worst thing to try to figure out how to use from its UNIX style manual page, but once you get it right for a given purpose you can put it in a script and do things automatically.

Sampling is largely a record-keeping exercise, and an exercise in planning and consistency. I use two Python tools I wrote to help with it (chopping a recorded velocity layer file into one file per sample, and building the sfz file from a little text configuration and the chopped audio files.) It's nerdware and not particularly easy to use the first time but perhaps I could post it if you're interested.

[GRUMP]

Sampling is largely a record-keeping exercise, and an exercise in planning and consistency.

I´d rather call it the "Royal Road". DAO.

I use two Python tools I wrote to help with it (chopping a recorded velocity layer file into one file per sample, and building the sfz file from a little text configuration and the chopped audio files.) It's nerdware and not particularly easy to use the first time but perhaps I could post it if you're interested.

Please don´t call me a Nerd now, but that sounds like just the right Tool for me. I create Multisamples every Day (Halion). Would you please ... be so kind?

[JeffLearman]

Please don´t call me a Nerd now, but that sounds like just the right Tool for me. I create Multisamples every Day (Halion). Would you please ... be so kind?

OK, give me a day or two.

The pitch detection was initially tweaked for Rhodes but works well enough for piano, except the very highest and lowest notes.

[hoanglongdummy]

Always sample the range of the original instrument, if there is an "original instrument." Don't limit yourself to C4-C7 unless that's the normal range for whatever you're sampling.

What are you sampling, btw? What format do you plan to use (sfz, kontakt, etc.)?

I'm currently now change to the new range C3 to C8, because I own a 61 keys piano.
+ Why record at below C3 or above C8 when the sound couldn't easily be heard,
+ 61 keys is something I can handle but 81 keys is a bit overboard (Most people own 49 or 61 keys midi controller, hell they even use computer keyboard)

I currently sampling my hardware and software synths, I maybe don't have to bother much with velocity and round robin much. However I do have to mind the mono and stereo thing.

I still wonder though do people dislike reverb in a patch that much ?

The format I aim at maybe SFZ and Kontakt only. (I don't have enough budget to buy Logic EXS24 or Steinberg Halion). There's another reason for this: I don't use MacOs so Kontakt may be a good choice for both Windows and Mac users.

Have you used this tool ? Could you tell me your thoughts ? It was Extreme Sample Converter or ESC
https://www.extranslator.com/

[JeffLearman]

If you're only doing this for your own use, then heck, only sample what you'll use! But if you plan to share with others, lots of people have 88-key boards, and lots of people create MIDI without a keyboard, or they transpose. In general (when not doing it just for yourself), sample whatever makes sense for the particular sample source. For many synths, you're right that a 61-key range is reasonable.

I do a lot of sampling only for myself. For example, I sample sounds from my digital piano just to make it easier to do DAW projects with those sounds. After the initial MIDI recording from my keyboard, I can edit the MIDI and re-render without attaching my piano, or without worrying about whether I touched the knobs on my audio interface since I recorded the part. And it renders to audio in seconds rather than having to re-play it through my keyboard, after any MIDI edit. Saves a ton of time. I can always re-render with the keyboard before a final mix, but I'll generally only do that for a solo piano piece, where the CP4's nuances outshine the sample set, or where I need half-pedaling.

BTW, terms like C3 aren't standard. Some software uses C4 as middle C. Other software calls it C3 or C5 and maybe even C6.

You don't need velocity layers unless the tone changes with velocity. Even then, in a lot of cases you can fake it just by modulating filter cutoff with velocity. For example, you can make a surprisingly decent piano by only sampling the highest velocity and lowering the cutoff for lower velocities.

I have used EXSC; that's my favorite tool for looping and some other stuff. It has a very quirky GUI, but I find it useful. Another tool to consider is Polyphone, which I haven't used but is held in high regard. But I wouldn't use EXSC to create a sample set, just to tweak it or convert formats.

[JeffLearman]

I would leave reverb out of the samples because reverb is a matter of taste and everyone has reverbs. But sometimes it might be crucial to what you're sampling.

Note that a reverb baked into a sample will stop immediately when the note stops, which is not how reverb normally sounds. The first time I samples a digital piano I forgot to turn the reverb off. I regretted that later! It didn't ruin it, but now when I use it I notice it and wish I'd thought of it. It just sounds unnatural, to have a moderately reverby piano sound but the reverb doesn't last after the notes. And adding more reverb causes too much reverb.

[GRUMP]

Please don´t call me a Nerd now, but that sounds like just the right Tool for me. I create Multisamples every Day (Halion). Would you please ... be so kind?

OK, give me a day or two.

The pitch detection was initially tweaked for Rhodes but works well enough for piano, except the very highest and lowest notes.

Pitch Detection is Rocket Science!

You have probably made it for Recordings of accoustic Instruments?

I´m curious to try it as soon as you´re ready (and the Code has proven the Security Tests!!)... my

How about an Alternative that slices the Source into a defined Number of equal sized Files - with the Option to define a File Naming Scheme for the Slices (1=xxx D1, 2=xxx D2, ... )?

I´ll maybe ask some Pro-Coders in my Network if you don´t expect it to make too much Effort?

Thanks in Advance and stay save!