KVR Audio

rmtek · Post by **rmtek** » Tue Mar 20, 2007 8:58 am

hello

I have some questions about multisampling:

1.) should i make one stereo files or separated left/right mono files?
2.) in 16bit/44,1khz or 24bit/48khz? (for me I take always 24bit/48khz)
3.) which notes/keys should i sample? (for me I take always each note(key of the octave)
4.) into which format should be exported? so that other useres could use it with there software? (for me i use SF2 soundfont format and *.sxt (Reason NNXT) that includes the samples and multisample parameter for keyzone,rootkey, loop etc)
5.) level? -0.3db, -0.6 Db or yet lower?
6.) if you got some other tipps i would be very proud.

To generate the samples use I presently:

1.) cubase + various vsti synths (to produce the samples)
2.) wavelab or soundforge (possibly rework of some samples)
3.) reason (to generate the multisamples with parameters in the NNXT)
4.) awave studio (to produce the SF2 file out of the NNXT *. sxt file
that awave studio can import, so that I get the parameters for keyzone,rootkey, loop etc)

greetz,
rmtek

p.s. please excuse my bad english

BertKoor · Post by **BertKoor** » Tue Mar 20, 2007 10:02 am

1) Whatever your sampler requires... If it's a mono sound source, keep it mono (one channel, not two) and don't bother with stereo at all.

2) 16bits should be sufficient. 44 or 48kHz doesn't make that much difference if the sampler has a good resampling algorythm. Sometimes it's safer to stay with lower rates, since then aliasing is less of a problem: it occurs mostly on downsampling, less on upsampling...

3) This depends on what you're sampling and what quality you're after. Sometimes four notes per octave is sufficient: only up/downsampling of one semi-tone is then required. Depending on the instrument you could also get away with two samples per octave. Size of the files can be important! Find the balance / trade-off that works for you...

4) SF2 and NNXT may be fine. Kontakt and Giga are popular also. Have a look what the big boys do if you really care.

5) Depends on the signal chain. If it's all digital (no microphones, preamps etc, just softsynths) then I'd use a peak level of 0.0dB

6) There are programs that can generate soundfonts (and other lib formats) from VSTi's. Just specify which notes, which length, which velocities and it does all the tedious work for you.

But why actually bother to create samples if you have the actual VSTi instruments? Samples always lack control, emotion, expressability and variation. LFO modulation for example is always made static.

If Reason is your main host/sequencer, then ReWire can be used to play VSTi's in other hosts. Maybe preservation of CPU power is your main concern, but then "freeze" is your friend.

Good luck!

rmtek · Post by **rmtek** » Tue Mar 20, 2007 11:42 am

thx for the fast responce!

6) There are programs that can generate soundfonts (and other lib formats) from VSTi's. Just specify which notes, which length, which velocities and it does all the tedious work for you.

did you know some of these tool by name?

thx for your help!

greetz

bobbyem · Post by **bobbyem** » Tue Mar 20, 2007 12:32 pm

[quote="rmtek"]hello
2.) in 16bit/44,1khz or 24bit/48khz? (for me I take always 24bit/48khz)
[/quote]

48khz is mostly for movies, better go with 44,1 less hassle later on

rmtek · Post by **rmtek** » Tue Mar 20, 2007 12:48 pm

hi,

ok, if 16bit/44,1khz is sufficient, then i will use that rate.

krushing · Post by **krushing** » Tue Mar 20, 2007 1:02 pm

6) There are programs that can generate soundfonts (and other lib formats) from VSTi's. Just specify which notes, which length, which velocities and it does all the tedious work for you.

did you know some of these tool by name?

I think HighLife does this. At least the specs say it exports "rgc:Audio SFZ" but I'm not entirely sure if this means they're actual .sf2 files then...

e// oh, Extreme Sample Converter seems to do this as well.

rmtek · Post by **rmtek** » Tue Mar 20, 2007 3:33 pm

big thx to you!

i will try the HighLife free plugin.

Jeremy_NSL · Post by **Jeremy_NSL** » Tue Mar 20, 2007 5:03 pm

rmtek wrote: 1.) should i make one stereo files or separated left/right mono files?

I can't think why you'd want dual-mono splits. Just regular stereo is fine, you can split later if you need to.

2.) in 16bit/44,1khz or 24bit/48khz? (for me I take always 24bit/48khz)

IMO 48khz is a waste of space. 24-bit is helpful though, even if you dither down to 16-bit later.

3.) which notes/keys should i sample? (for me I take always each note(key of the octave)

To me thats mostly overkill. 3-4 notes per octave are generally ok. The more notes, the more looping/processing work.

4.) into which format should be exported? so that other useres could use it with there software? (for me i use SF2 soundfont format and *.sxt (Reason NNXT) that includes the samples and multisample parameter for keyzone,rootkey, loop etc)

I map in sfz which is easily convertable to my other target formats.

5.) level? -0.3db, -0.6 Db or yet lower?

I normalize to -0.05 db because Rene @ Cakewalk told me to

Mostly its not a big difference anyway.

Random tip: if you are sampling synth stuff, its generally best to remove most modulations ie. lfo stuff, chorus, delay & reverb. These are hard to loop and can usually be duplicated in a sampler/rompler later.

Good luck!

mahtazz · Post by **mahtazz** » Tue Mar 20, 2007 5:05 pm

Great tips.

trixie · Post by **trixie** » Tue Mar 20, 2007 5:31 pm

krushing wrote:At least the specs say it exports "rgc:Audio SFZ" but I'm not entirely sure if this means they're actual .sf2 files then...

SFZ is NOT SoundFont (.sf2). It is an open sampler format designed by Rene for his SFZ player. It's becoming increasingly popular and is now supported by some other samplers.

rmtek · Post by **rmtek** » Wed Mar 21, 2007 7:29 am

yes and with awave studio could .sfz be exportet in .sf2 or others.

rmtek · Post by **rmtek** » Wed Mar 21, 2007 3:52 pm

hi,

are there some other plugins/tool like HighLife?

gumbo · Post by **gumbo** » Fri Mar 23, 2007 9:58 pm

rmtek wrote:hi,

are there some other plugins/tool like HighLife?

Have you checked out SampleRobot?

learjeff · Post by **learjeff** » Tue Mar 27, 2007 6:55 pm

I recommend recording in 24 bits, processing as needed, then dither to 16 bits before building the result. However, if you're not normalizing each velocity layer, stay in 24 bits for the whole process. (There are other adjustments you need to make if you're not normalizing each velocity layer.)

I have some python scripts that will chop layer files into individual samples (labeled according to pitch / midi note, which is automatically detected), build a keyboard map and sfz file from the results, and finally build an sf2 file as optional last step. For more info, see http://learjeff.net/sf/sf.html and follow the 'tools' link.

However, if you're sampling a vsti, it's easier and better to simply use one of the various vsti capture tools. Sorry, haven't used one, but this is clearly the way to go.

rmtek · Post by **rmtek** » Wed Mar 28, 2007 1:42 pm

hello,

big thx for your help!

multisampling questions