Tag: sample library

Posted on by — Leave a comment

For Sample Creators: General Advice on Making Sample Libraries Based on Real-World Instruments

Hello Samplists,

Here’s a list of general advice for people making sample libraries based on real-world instruments. All of this advice falls squarely into the “give advice that you yourself need to hear” category. I should also mention that this is an intermediate-level post that assumes that you already know that basics of how to make a sample library.

1. Don’t get hung up on capturing the “authentic” sound of an instrument

Think carefully about your actual goal when making a sampled instrument. When I first started, I approached the work of making sample libraries as though I was an archivist trying to preserve the “authentic” sound of a real world instrument for posterity. This is a noble goal, but it doesn’t necessarily yield the best sampled instruments. I quickly realized a much better goal was trying to make virtual instruments that are fun to play, sound good, and useful to composers or producers. And yes, sometimes these two goals are at odds with each other. 

For example, when you record a piano, you may discover that some notes sound great and others sound lousy. If you are trying to be accurate, you may think you should include those lousy notes in the same, because, after all, they also represent how the instrument truly sounds. Of course, this is a valid perspective. I would argue that you should leave them out, or at very least, provide a version with a curated selection of “good” notes. The vast majority of composers and producers would rather have something that just sounds good immediately, that they don’t have to wrestle a good sound out of.

Another important thing to realize is that samplists are never getting the “true” sound of an instrument anyway. When it comes to acoustic instruments, even something as basic as where you put your microphone can drastically change the way an instrument sounds. In other words, even someone trying to capture the authentic nature of instrument is still making creative and aesthetic decisions. So give that you’re already inserting yourself into the process at the recording stage, why not continue making aesthetic decisions at every point in the process. If you record a violin sample, and it sounds harsh and grating, by all means EQ it until it sounds good! If it sounds to dry, add some reverb.

I’ve made the mistake several times of faithfully re-creating, an instrument sound, only to find that for whatever reason that sound didn’t actually work as a sample library. It’s been helpful for me to think of the source sound as the departure point on which I’m building a brand new Instrument – a virtual Instrument – that will be most likely be triggered using a piano keyboard.

2. You don’t need nearly the number of samples you think you do. 

When most people are getting started, they see sample libraries releases by big sample library companies boasting about how many samples are included in a release and they will naturally think that they also need to record every note of an instrument with five different velocities and sometimes even round robins. I’m here to tell you that 9 times out of 10 this is not necessary. Most users would rather save the hard disk space. Most melodic instruments you can have a zone every 3 to 6 notes — sometimes even just one zone per octave – and it will sound every bit as good. 

As you get more experienced, you will discover which instruments require more samples and which can do without. For example, one exception is when there is any sort of modulation like tremolo, vibrato or an LFO filter on a synth. If you try to pitch bend samples that have these sorts of fast, time-based modulation, you will end up with different rates of modulation depending on which note you play, which may sound terrible. In other words, use your aesthetic judgement.

3. Always pitch-bend down

If have not recorded a sample for every note, you may need to set up mapping so that pitch-bending takes place. Often, when you map samples in a sample-mapping product (such as Logic Sampler or Kontakt), the mapper will want to put the root note in the middle of a zone, meaning that if you play a play a note above the root note, it will pitchblende the note up. For example:

A screenshot of the Logic sampler's default mapping.

Here’s the thing: samples that are pitch-bent down almost always sound better than samples that are pitch bent up. This is because when you pitch-bend samples down, you remove some of the higher frequencies, which at worst just sounds a bit like a nice low-pass filter. On the other hand, when you pitch-bend a sample up, you are actually adding high frequencies, which can sound weird and “chipmunky.”

The solution is to always map your samples with the root note at the top of the range. To continue our example from above, you would get something that looks like this:

A more desirably sample mapping

4. How to know when your instrument is done and ready to release

This is easy. Your sample is done when you can’t stop playing with it. When every time you open it up the library, you get distracted and start writing music. You are your own first customer, so be honest with yourself. If your own reaction to playing with a library is a tepid “I guess this is good?” then there’s something wrong. Figure out what’s bothering you about the sounds and fix them.

Hope this little list has been helpful. As I think of more things, I will go back and update this document as more things occur to me.

All the best,
Dave

Posted on by — 11 Comments

For Sample Creators: How to use Convolution in your Decent Sampler presets

A spectrogram of a convolution reverb impulse response.

Version 1.6.12 of Decent Sampler brings a Convolution effect to the Decent Sampler platform. If you don’t know what Convolution is, you can see a great explanation here. The most common use case for convolution is in creating reverb, and that is the use case that will be demonstrated here.

How to add the Convolution effect to a preset

The convolution effect is invoked in much the same way that any other effect is defined:

<effects>
  <effect type="convolution" mix="0.5" irFile="Samples/Hall_IR.wav" />
</effects>Code language: HTML, XML (xml)

As you can see, other than the required type attribute, there are two other attributes:

  • The mix attribute controls how much of the convolved signal is present in the output. A value of 0 is completely dry whereas a value of 1 is completely wet containing only the convolved signal.
  • The irFile attribute specifies the file that should be used as an impulse response or IR.

How to control the convolution effect using UI controls

Two of the convolution effect’s attributes can be controlled using UI controls. The mix level can be controlled by a knob as follows:

<labeled-knob x="680" y="40" label="Conv Mix" type="float" minValue="0" maxValue="1" value="0.5" textColor="FF000000" >
  <binding type="effect" level="instrument" position="0" parameter="FX_MIX" translation="linear"  />
</labeled-knob>Code language: HTML, XML (xml)

The IR impulse can be changed dynamically using a menu control:

<label text="IR File" x="480" y="40" width="120" height="30"></label>
<menu x="580" y="40"  width="120" height="30" requireSelection="true" placeholderText="Choose..." value="1">
  <option name="long hall.wav">
    <binding type="effect" level="instrument" position="1" parameter="FX_IR_FILE" translation="fixed_value" translationValue="Samples/long hall.wav" />
  </option>
  <option name="ABLCR Chord Vocal.aif">
   <binding type="effect" level="instrument" position="1" parameter="FX_IR_FILE" translation="fixed_value" translationValue="Samples/ABLCR Chord Vocal.aif" />
  </option>
  <option name="Amp Spring High.aif">
    <binding type="effect" level="instrument" position="1" parameter="FX_IR_FILE" translation="fixed_value" translationValue="Samples/Amp Spring High.aif" />
  </option>
  <option name="Swede Plate 3.5s.aif">
    <binding type="effect" level="instrument" position="1" parameter="FX_IR_FILE" translation="fixed_value" translationValue="Samples/Swede Plate 3.5s.aif" />
  </option>
</menu>Code language: HTML, XML (xml)

Examples

An example Decent Sampler preset that uses IR reverb can be downloaded here.

Performance considerations

While convolution is a powerful tool that can go a long way towards shaping a sample library’s sound, it can also be quite costly in terms of CPU usage. Sample creators would do well to create versions both with and without convolution effect and compare the relative CPU usage of the two versions before opting to use convolution.

Posted on by — 10 Comments

How to add LFOs and extra envelopes to your Decent Sampler instruments

As for version 1.5.24 of Decent Sampler, it is now possible to make use of LFOs and ADSR envelopes in your Decent Sampler sample libraries. In this blog post, we’ll go through how to set these up.

The <modulators> section

This is a new section that lives below the top-level <DecentSampler> element and it is where all modulators for the entire sample library live:

<DecentSampler>
    <modulators>
        <!-- Your modulators go here. -->
    </modulators>
</DecentSampler>

The <lfo> element

Underneath the <modulators> section, you can have any number of different LFOs, which are defined using an <lfo> element, for example:

<modulators>1
  <lfo shape="sine" frequency="2" modAmount="1.0"></lfo>
</modulators>

This element has the following attributes:

  • shape: controls the oscillator shape. Possible values are sine, square, saw.
  • frequency: The speed of the LFO in cycles per second. For example, a value of 10 would mean that the waveform repeats ten times per second.
  • modAmount: This value between 0 and 1 controls how much the modulation affects the things it is targeting. In conventional terms, this is like the modulation depth. Default value: 1.0.
  • scope: Whether or not this LFO exists for all notes or whether each keypress gets its own LFO. Possible values are global (default for LFOs) and voice. If voice is chosen, a new LFO is started each time a new note is pressed.

The <envelope> element

In addition to LFOs, you can also have additional ADSR envelopes. These can be useful for controlling group-level effects, such as low-pass filters. If this is what you wish to achieve, make sure you check out the section on group-level effects below.

To create an envelope, use an <envelope> element:

This element has the following attributes:

  • attack: The length in seconds of the attack portion of the ADSR envelope
  • decay: The length in seconds of the decay portion of the ADSR envelope
  • sustain: The height of the sustain portion of the ADSR envelope. This is expressed as a value between 0 and 1.
  • release: The length in seconds of the release portion of the ADSR envelope
  • modAmount: This value between 0 and 1 controls how much the modulation affects the things it is targeting. In conventional terms, this is like the modulation depth. Default value: 1.0.
  • scope: Whether or not this LFO exists for all notes or whether each keypress gets its own LFO. Possible values are global and voice (default for envelopes). If voice is chosen, a new LFO is started each time a new note is pressed.

How to use bindings in conjunction with modulators

In order to actually have your LFOs and envelopes do anything, you need to have bindings under them. If you are not familiar with the concept of bindings, you may want to read this section of the File Format Reference Guide and then return here. Bindings tell the engine which parameters the LFO should be affecting and how. Here is an example:

<modulators>
    <lfo shape="sine" frequency="2" modAmount="1.0">
        <!-- This binding modifies the frequency of a low-pass filter  -->
        <binding type="effect" level="instrument" effectIndex="0" parameter="FX_FILTER_FREQUENCY" modBehavior="add" translation="linear" translationOutputMin="0" translationOutputMax="2000.0"  />
    </lfo>
</modulators>

This is Example 1 from the example pack you can download here.

How modulator bindings differ from knob or MIDI bindings

If you’re already familiar with the concept of bindings, then you’ll want to read this section especially careful as you may notice a few difference between bindings as they are used by knobs and the ones used by modulators. Specifically, when you move a UI control that has a binding attached, the engine actually goes out and changes the value of the parameter that is targeted by that binding. For example, if you have a knob that controls a lowpass filter’s cutoff frequency, moving that knob will cause that actual frequency of that filter to change. In other words, the changes that the knob is making on the underlying sample library are permanent. The same is also true for bindings associated with MIDI continuous controllers.

Modulators, on the other hand, do not work this way. If a modulator (such as an LFO) changes its value, the engine looks at the bindings associated with that LFO and then makes a list of temporary changes to the underlying data. When it comes time to render out the effect, it consults both the permanent value and the temporary modulation values. As a result of this difference in the way bindings are handled, only some parameters are “modulatable.” At time of press, the following parameters are modulatable:

  • All gain effect parameters
  • All delay effect parameters
  • All phaser effect parameters
  • All filter effect parameters
  • All reverb effect parameters
  • All chorus effect parameters
  • Group Volume
  • Global Volume
  • Group Pan
  • Global Pan
  • Group Tuning
  • Global Tuning

The new modBehavior parameter for bindings

Another new feature of bindings is the addition of the modBehavior attribute. This controls exactly what effect a binding actually has on the parameter it is targeting. There are three possible values for this:

  • set: This means that the value that is generated by the binding becomes the new value for the parameter being targeted. NOTE: set is the default value and this is the way that knobs and MIDI CC bindings work by default. That being said, it’s usually not the correct choice for modulations such as LFOs and secondary ADSR envelopes.
  • add: The value generated by the binding gets added to the current value of the parameter being targeted.
  • multiply: The value generated by the binding gets multiplied with the current value of the parameter being targeted.

In order to understand what any of this means, let’s look at the following example:

<effects>
    <effect type="lowpass" frequency="60.0"/>
</effects>
<modulators>
    <lfo shape="sine" frequency="2" modAmount="1.0">
        <binding type="effect" level="instrument" position="0" parameter="FX_FILTER_FREQUENCY" translation="linear" translationOutputMin="0" translationOutputMax="2000.0" modBehavior="add" />
    </lfo>
</modulators>

The <lfo> tag above sets up an LFO with a frequency of 2 Hz. It has just one binding, which targets the first global effect, which happens to be a low-pass filter with a cutoff frequency of 60Hz. Every binding can be seen as a pipe that takes an input value, translates that value in some fashion, and then sets a parameter somewhere else in the engine. Here are the steps for this setup:

  1. By default, an LFO generates values between -1.0 and 1.0.
  2. These values then get passed to the binding, which is setup to do a linear translation. This linear translation has a minimum of 0 and maximum of 2000, which means that when the LFO is at its lowest point (-1.0) the binding will generate the number 0 (the minimum) and when the LFO is at its highest point (1.0), the binding will generate the number 2000 (the maximum).
  3. Because the modBehavior value is add, this new value that is generated by the binding will be added to the original cutoff value of 60Hz. This means that when the LFO is at its lowest point, the filter cutoff will be 60Hz (i.e. 60 + 0) and when its at its highest point, the filter cutoff will be 2060Hz (i.e. 60 + 2000).

LFO Scope: Global or Voice-level

By default, all modulators will be created at the global level. This means that there will be exactly one modulator that is shared by all voices. In many situations, such as an LFO modulating a single low-pass filter which is shared by all of voices, this is often what we want.

But there are other situations where we don’t want our modulator to be global. For example, what if we want to have an envelope that targets a low-pass filter. Let’s say that when we press down on a key, we want that low-pass filter to open up slowly until, 2 seconds later, it reaches its peak. In theory, we could set up something like this:

 

<effects>
    <effect type="lowpass" frequency="60.0"/>
  </effects>
<modulators>
    <envelope attack="2" decay="0" sustain="1" release="0.5" modAmount="1.0">
        <binding type="effect" level="instrument" position="0" parameter="FX_FILTER_FREQUENCY" translation="linear" translationOutputMin="0" translationOutputMax="4000.0" modBehavior="add" />
    </envelope>
</modulators>

But there’s a problem with this: let’s imagine that we hit a note, and then one second into that first note, we hit another note. If we have just a single envelope, that envelope will be half-way through its attack phase when the second key is pressed. Depending on how the envelope is configured, that envelope will either retrigger because of the new keypress (this would be the wrong behavior for the first note which is still being held) or keep going in which case the second note will start half way through its attack phase.

To solve this problem, in such cases, we need tell the engine to create a separate modulator for every keypress. To do this we add a scope="voice" attribute to the modulator as follows:

<envelope attack="2" decay="0" sustain="1" release="0.5" modAmount="1.0" scope="voice">

But wait, there’s another problem! In the scenario above, even if we have separate modulator for every keypress, those voices are still all sharing a single global low-pass filter. If you’ve got several modulators pinned to the same global effect, they are going to be setting and resetting that global effect’s parameters to competing values. The engine is going to be at war with itself of which envelope’s values are the correct setting for the filter’s cutoff. In other words, we need separate effects for each keypress. These can be added by specifying…

…effects at the group level

Adding effects that only apply to a specific group is easy. All you need to do is create an <effects> group that lives underneath the <group> element for the group you want to affect. For example:

<groups>
    <group>
        <!-- A sample -->
        <sample path="Samples/Volca Keys Poly-V127-60-C3.wav" loNote="10" hiNote="83" rootNote="48"/>
        <effects>
            <!-- These effects will only apply to this group -->
            <effect type="lowpass" frequency="22000.0"/>
        </effects>
    </group>
</groups>

Group level effects are initialized every time a note is started and destroyed every time a note is stopped. If you play two notes simultaneously, two instances of this effect will be created and these will be independent of eachother. As a result, they use more CPU than global effects.

NOTE: Only certain effects will work as group-level effects: lowpass filter, hipass filter, bandpass filter, gain, and chorus. Delay and reverb cannot work properly as they will be deleted before their tail peters out.

How to have knobs control group-level effects

Just as it is possible to have knobs that control instrument-level effects, it is also possible to have them control group level effects. In order to specify a group level effect, set the binding’s level to group, and use groupIndex and effectIndex to specify which specific effect needs to be controlled. Here is an example of this a knob that controls a group-level low-pass filter:

<labeled-knob x="655" y="75" label="Tone" type="float" minValue="0" maxValue="1" value="1">
    <binding type="effect" level="group" groupIndex="0" effectIndex="0" parameter="FX_FILTER_FREQUENCY" translation="table" translationTable="0,33;0.3,150;0.4,450;0.5,1100;0.7,4100;0.9,11000;1.0001,22000"/>
</labeled-knob>

Putting it all together: an envelope filter that controls a low-pass filter

So, if we put all of this into a real world example, we can code imagine an ADSR envelope that is controlling a low-pass filter as follows:

<?xml version="1.0" encoding="UTF-8"?>
<DecentSampler minVersion="1.6.0">
  <groups>
    <group ampVelTrack="0.0">
      <sample path="Samples/Volca Keys Poly-V127-60-C3.wav" loNote="10" hiNote="83" rootNote="48"/>
      <effects>
        <effect type="lowpass" frequency="22000.0"/>
      </effects>
    </group>
  </groups>
  <modulators>
    <envelope attack="1" decay="0.5" sustain="0" release="0.5" modAmount="1" scope="voice">
      <binding type="effect" level="group" groupIndex="0" effectIndex="0" parameter="FX_FILTER_FREQUENCY" modBehavior="set" translation="table" translationTable="0,33;0.3,150;0.4,450;0.5,1100;0.7,4100;0.9,11000;1.0001,22000"  />
    </envelope>
  </modulators>
</DecentSampler>

This is Example 2 from the example pack.

How to make knobs control modulator parameters

Both types of modulators – <lfo> and <envelope> – have a modAmount parameter. This consists of a value between 0 and 1 that dictates how much the modulation affects the things it is targeting. In conventional terms, this is like the modulation depth. In order to create a knob that controls LFO depth, you would create a knob that targets a modulator’s depth. Here’s how you would do this:

<labeled-knob x="585" y="75" label="LFO Depth" type="float" minValue="0.0" maxValue="1" value="1" >
    <binding level="instrument" type="modulator" position="0" parameter="MOD_AMOUNT" />
</labeled-knob>

Note the binding type value of modulator and a position of 0. In other words, we are targeting the first modulator in the modulator block. The parameter we are targeting is MOD_AMOUNT.

It is similarly possible to target LFO rate using a parameter value of FREQUENCY:

<labeled-knob x="655" y="75" label="LFO Rate" type="float" minValue="0.0" maxValue="10" value="1">
   <binding level="instrument" type="modulator" position="0" parameter="FREQUENCY" />
</labeled-knob>Code language: HTML, XML (xml)

You can see a full example of a patch that controls rate and depth in Example 3 of the example pack.

Conclusion

That’s pretty much it. We look forward to seeing what you do with LFOs and envelopes. Make sure you download the example pack from here.

The example pack contains the following examples:

  • Example 1 shows how to make an LFO that controls a global lowpass filter
  • Example 2 shows how to have a voice-level envelope which controls a group-level filter
  • Example 3 shows how to have knobs control LFO parameters
  • Example 4 shows how to have knobs control envelope parameters
  • Example 5 shows how to modulate group volumes
  • Example 6 shows how to modulate group panning
  • Example 7 shows how to modulate group tuning