Making multi-samples

[Dave N]

When you make a multi-sample, do you normalize each sampled note or the whole set of notes as a whole to maintain the level differences between sampled notes. The way i do it now is to record the sequence of notes and then normalize the whole file after which i add markers and export the individual samples. This often has the effect that some notes are louder than others keeping a lot of the other samples kind of soft.

Hope this makes any sense

Looking forward to a reply. BTW: Great community!

[Failed Muso]

I normally sample each note individually and then either normalise individually within the Akai or batch normalise using Adobe Audition.

Given your method, I'd slice up the notes and then either normalise individually or batch process them.

[Dave N]

Ahh so you do normalize each note individually. Thanks for the reply, this helps me a lot!

[Failed Muso]

Ahh so you do normalize each note individually. Thanks for the reply, this helps me a lot!

No probs

Just make sure you maintain the same velocity and then normalize each one either individually or as part of a batch process and you should be ok.

[smoothjay]

i usually normalize the set of samples together to maintain the ratio of volume amongst them.

But I have seen it done both ways and you can get decent results either way.

The problem I have come across is that the lower level articulations (the soft performances) tend to have more noise, so normalizing them as single hits can raise the noise floor a little too high.

[Jake Jackson]

The goal is to avoid normalizing entirely. May be impossible, but to get a realistic set of samples, one should ideally record in a noiseless room with a very low sound floor (excellent equipment and a low signal to noise level, in other words): you want to record the actual volume of each note.

This means that many soft layers will be needed to get the best results.

[Rangtangtang]

If we are sampling a keyboard, well I think, you'd normalize them all as a set, if you want it to sound more like the original? That is if you doing every note?

[Dave N]

I'm multi-sampling synths. I think a synth doesn't really have the fixed volume ratio amongst the notes as each patch shows me different volumes in notes across the keyboard. One patch might have most notes being almost identical in volume while the other has various regions being louder than others. Many times you have like 2 notes which are louder than the others making the general volume of the patch quite low. I think normalizing each note (not too much to avoid noise) is the way to go indeed. It also doesn't really matter with synths as they don't have a fixed sound to start with in the first place. If it would be acoustic instruments it would be a different story altogether

Thanx for the feedback guys.

[Dave N]

Just did a little experiment with a soundset which i normalized each individual note from. Loaded the set into the sampler (NN-XT) and exported the notes from Reason. Seems like the sampler added a new volume ratio to the set with some notes being softer and others being louder again. Anyone any thoughts on this?

[pljones]

It depends on a number of things.

One of them is the sampler you're going to use for playback. SFZ expects each sample to be normalised (within a keygroup). (By default) it uses velocity to scale the playback volume of the sample. If you haven't normalised the samples, you need to adjust the default scaling to take that into account.

Of course, there's more to it than volume. Across keygroups, particularly, filters may be more or less open (even at the same velocity) and all that needs tidying up in the mapping process, if you can.

[Sascha Franck]

It all really depends.
Usually I slap the sampled notes into my sampler of choice first, without any destructive normalizing, to see whether things will be fine or not.

If you have a whole bunch of velocity zones, you may better be off not normalizing them at all, thus maintaining the original relationship of the samples. In that case you'd have to set your sampler not to route velocity to volume, as that'd result in a doubled velocity response.

If you however have just very few zones, normalizing might work out better, letting the sampler do the volume scaling.

It still might depend though. As said, the best way would be to slap the zones onto your sampler before doing anything, just to give things a try.

[Dave N]

The samples which i make don't have multiple velocity zones. They are recorded at a single velocity. I'm not really sampling a synths respons but more making a bank of sounds which can then be used as layers to create new sounds in the sampler. This will be a collection which i might release commercially one day so i would like it to be as usefull as possible therefor this question

[Sascha Franck]

Allright, then try to get their level as hot as possible during recording.
Normalizing shouldn't be required anyways as it'd only mean the noise floor would be raised as well. Simple volume adjustments can be done on the sampler itself.

[soniccouture]

Have to disagree. One should always normalise then use the sampler to do anyvelocity/ keyboard amplitude scaling. It gets a more consitent result once you apply volume scaling - it'll be smoother, and you'll have more control, particularly if you use the samples in other patches later.

[pljones]

Um, what are you disagreeing with? If the original samples are as hot as possible, normalisation will do nothing (apart from maybe degrade the quality). Sacha also says that the sampler should do the simple scaling adjustments you're talking about.