Who is the best sample-based DJ

Sampling Tutorial: This is how you can create your own samples

Workshop: create sample banks yourself (recording)

DIY: Our tips for sampling

When it comes to sampling, opinions differ. For some it is a blessing, others say it is used to pirate the sound of a musician or an instrument. However, you can use this technology for good if you use it creatively. Comparable to a video recorder with which you could watch the recorded TV program over and over again, a sampler offers you the opportunity to play sounds over and over again that are not permanent or that come from instruments that were only available to you for a short time - right now long enough to sample. In the following workshop you can read about how easy this is basically and what you should pay attention to when sampling.

In contrast to today's digital sampler instruments, the first samplers in music history were hardware-based. Today we have software samplers that can be perfectly integrated into the DAW and make the world of sampling easy for everyone. You no longer have to be a sampling specialist to create your own sample bank. So why not build your own, very personal sample bank? Basically everyone who has a DAW, an audio interface and a microphone at the start has everything they need at their disposal.


Sampling can be broken down into three phases: planning, recording, and editing. In the planning you have to weigh the technical effort, the recording itself has to take place under certain conditions and in the editing the playback parameters of the samples are determined. Of course, a lot of it is comparable to conventional recording of music or sounds. In detail, however, there are a few crucial tips to ensure that the sample bank sounds good and is easy to use.

What do I have to consider if I want to quickly sample a sound from the borrowed Prodigy and what is important if I want to produce a more complex multi-sample bank? So let's separate the sample chaff from the wheat ...


1. The planning

The right place

Sampling an instrument with audio outputs requires the least amount of planning. Both the spatial and the technical requirements are minimal, so that in principle you only have to set up the necessary cabling for the audio signal chain to get started. If the sound generator is an instrument that has no audio outputs and one or more microphones have to be used, one of the most important decisions is to choose a suitable location for the recordings. In contrast to conventional recordings, in which a recorded background noise is only heard once at a time, namely when playing back that point, a background noise in a sample would sound again and again with every keystroke or triggering of the sample and so in the long run annoy and make the sample unusable. When sampling with microphones, it is important to choose a location that is as quiet and quiet as possible.

The player

Someone has to make the instrument or body of sound to be sampled sound. There are three possibilities here: You either play yourself, program a MIDI track or a second person acts as a player who should be chosen particularly wisely, as you have to be particularly quiet and disciplined when sampling with microphones to avoid the subsequent samples To fill in extraneous noises such as breathing too loudly. During my many years of activity in the field of sampling, similar statements came up again and again in conversations with the musicians: Sampling is a bit like a meditation for the player, in which the musician can mainly practice his physical discipline on the instrument. For quiet, close-mic sampling recordings, it is important to be particularly quiet when playing. Even clothing that audibly rubs when moving or stomach growling can become a nerve factor in every sampling session over time. So don't choose a fidgety person for the sampling recordings and watch out for quiet clothes.

Technical effort

When sampling a synthesizer using line cabling, one cannot really speak of effort. Into the audio interface - and that's it! It looks different if you want to sample different positions in the room or even want to capture a unique constellation that you couldn't get organized a second time in a hurry. Like a choir of several singers or unique sound events like smashing a bottle or whatever else comes to mind. On such more unique occasions, it makes sense to set up as many microphones as possible and use particularly high-quality, good-sounding, low-noise technology. As soon as such a particularly complex sampling session is due, you should also think about a backup system, recordings on several hard drives and safeguarding the power supply. A UPS (uninterruptible power supply), which lets you save at least the recordings made up to that point on the hard drive in the event of a voltage drop, can save a lot of money and nerves for the player.


Sure: you use suitable microphones for sampling various acoustic instruments and amps / cabinets.

Suitable instruments

Of course, you can sample anything that produces a sound in any way, but when you have a choice of sound generators or instruments you should take the time to find out a few things. Is a certain instrument preferable because it might keep the mood better or because there is little or no background noise? Of course you can also start sampling straight away, but with synthesizers a sound is often alive because of the filters. If you were to sample a sound with a dynamic filter course, you would freeze the otherwise variably filtered sound virtually digitally and later play the same filter course again and again in the sampler. If a synth sound lives from its filter, it would be worth considering switching this filter off, i.e. only recording the underlying sound and later emulating the filter within the sampler engine. Since the filters of the two systems are guaranteed to differ in sound, the exact same sound would not have been available on the sampler, but the liveliness of the sound would have been preserved that way.

2. The recording

Location check

When sampling with microphones, it is particularly important not to capture any interfering noises from the room or the environment. As with conventional recording, checks the recording location particularly carefully for background noises such as traffic noise, loud neighbors, flies in the room and so on. This is particularly important because any noise would sound over and over again with each new playback of the later sample.

Dynamic levels

A good sounding and playable sample bank requires extra effort in making the underlying recordings. Preventing background noise is only half the battle. When sampling several volumes, it is important that the loudness of the individual tones per dynamic level sounds as homogeneous as possible: a sound that does not change in different volumes except in its loudness or a synthesizer preset that you only want to sample at full volume anyway. is easiest to capture. You record a loud or the loudest volume and can later play the finished bank at this one volume. Or you can program a velocity-sensitive volume playback parameter generated by the sampler, the so-called velocity function. If a sound body has particularly large sound differences when playing at different volumes, as many dynamic levels as possible should be recorded. The more dynamic levels you capture, the more authentic and true to the original the sound behavior of the original instrument or body of sound can be reproduced in the sampler. However, experience shows that, for example, a piano with twenty dynamic levels does not sound much more authentic afterwards than one with only eight or ten. A well-coordinated cross thread between a few dynamic levels can save a lot of sampling work here.

Here you can see: There are a lot of dynamic levels that were used for sampling.

My many years of experience in the production of sample libraries has shown that for most keyboard instruments four to six volume layers are sufficient to achieve an authentic sound progression through the volumes. Of course, this depends on the sound differences that can be extracted from the instrument. For a rocked piano, four to five layers are certainly sufficient if you want to make an effort. If you want to sample a particularly dynamic sounding grand piano, it can be worthwhile to record eight to ten layers. With orchestral instruments you get very good sample sets with four to eight layers. In any case, you should test the dynamics of the sound body to be recorded before sampling in order to find out how big the sound differences are and how many dynamic levels are necessary.

Ultimately, however, you are not only limited by the sound properties of an instrument, but also by the skills of the player. When it comes to time-consuming sampling, it is better to record too much than too little. Samples can still be left out afterwards, but recording them afterwards is usually very time-consuming, because when you repeat the recordings you have to restore all the conditions so that the sample bank sounds homogeneous.


How many tones of a scale are necessary depends on the one hand on the capabilities of the sampler engine and on the other hand on the sound properties of the source material. The current software samplers sound much better than they did ten years ago, so that you don't necessarily have to sample every semitone of an instrument from the sampler side in order to achieve a uniform sound. Whether you can sample not chromatically, but only in whole-tone steps or thirds, depends above all on the sound itself. If there are regular beats like a vibrato or a tremolo, it is imperative that you record the entire range in all semitone steps. If this is not done, the corresponding frozen modulations will be faster or slower depending on the mapping of the key zones when pitching in the sampler engine later. If you have vibrato or tremolo, you have to bite the bullet and record all semitone steps. It is similar with sounds with strong formants. If you record too few pitch levels here, the subsequent sample bank will sound unbalanced and artificial. Recording in whole steps is a good compromise to be on the safe side. So you don't have to record every single note and still have enough notes without the finished sample sounding artificial. Here, too, the tip: Better to take in too much than too little. In order to get the sample bank as small as possible in the end, when programming the bank you can always check what it sounds like if you leave out samples.

Tone length

The required length of a sample basically depends on whether it is either a short, percussive sound - such as a snare drum - or a long-lasting sound. With short, percussive sounds, it is often enough to record 1 second, depending on the decay behavior of the sound body and the room. With lively, long sounds, you don't necessarily have to record for minutes. Long notes can be played as loops in the sampler as often and for a long time. Depending on the sound, you need between 5 and 10 seconds and in most cases you should be able to inaudibly cross-fade the loop points with a 200-2000 millisecond crossfade. With powerful, more complex group sounds, it is easier to create inaudible loops than with fine, more fragile soloist tones. Also at this point the tip: longer recording is better than too short. Given today's low prices for hard drives, it would be nonsense to record samples that are too short because of capacity concerns.

Long fade-outs should also be recorded.

Variations of the game types

When it comes to a synthesizer sound, a piano, or a bass, there aren't too many variations. The decision becomes more difficult with string instruments or with the human voice. A violin can offer a great variety of playing styles, in which one strokes, strikes, brushes, bounces with the bow and the expressions of the human voice offer a great variety anyway. As a starting point, it is sufficient to first decide on a good-sounding set of long notes. Thanks to the ADSR curves of the sampler engine, long samples can not only be played long but also short and so you kill two birds with one sampling set. Ultimately it is only a question of the time available and sometimes also of the budget how many different types of game you can implement.

Repetitions per sample

In order to give the later sample bank a natural sound, it is important to have several versions of the same sound available. To avoid the so-called machine gun effect, one should have at least two repetitions of a sample. If you trigger the same sample several times in a row, you can immediately hear that it is a retort sound. This facade crumbles from two samples. The more repetitions available, the more natural the sample set will sound. If you have four or more samples, you are usually on the safe side. When four repetitions are played in random order with the help of a clever sampler engine, not everyone immediately realizes that it is a sampler and not the real instrument or the underlying sound body.

It is important that the recorded repetitions sound as similar as possible. This may sound contradicting at first, but it is easy to explain: For example, if a sample of four repetitions has a clearly audible peculiarity in the sound, after several runs of the sample you will eventually hear a pattern and thus lose the illusion of a real body of sound. Such an audible peculiarity then sounds like an accent with repetitive triggering that you would hear more and more easily over time. So make sure that the repetitions sound as similar as possible to your ears. Even precisely uniformly played piano notes have enough uniqueness at the waveform level that they can never be identical. With guitars, vocals or analog synthesizers, despite all the efforts, one tone will never be identical to another. The only exception: If you want to sample a sound from a sample-based or synced digital oscillator without any additional effects, it is very likely that you will get identical samples with precise play. Here it is sufficient to record a single variation and prevent a machine gun with the help of effects or modulations within the sampler engine.

Record some music too

In order to be able to better assess the sound properties of the instrument, the room and the microphones and, above all, their position and combination, one should record a short musical passage before the many individual tones. The easiest way to find out which microphone sounds best or which combination of spot and room mics might be ideal is with music. In the case of individual tones, it is increasingly difficult to assess the overall sound of a sample set, whereas one is more used to listening to music and thus quickly gets an impression of whether the location in the room is perhaps unsuitable or whether strange comb filter effects occur in combination with certain microphones. This is easier to recognize in musical passages and can possibly react by changing the microphone positions.

3. Editing


The post-processing of samples differs in only two things from the post-processing of conventional recordings, namely in so-called tagging and denoising.

The tagging (German: labeling) is a tolerable hard work, which can be a bit slow, especially with complex sample sets with several hundred samples. If I have recorded ten snare samples, the minimalist may be satisfied with the crude numbering from one to ten. However, if you have recorded several volume layers of a violin with many repetitions and articulations, clear tagging is not only clearer, but also makes it easier to correctly compile and program the sample set. Where in the past, due to the operating system, you mostly had to reduce yourself to eight-digit file names and therefore had to come up with clever abbreviations, you are not too limited with today's computer systems. What used to be called “VnC4pzf1.wav” can now be called “Violine_C4_pizzicato_forte__AKGC414_roommic_01.wav”. So that a sample can be assigned to a specific key, the sampler engine needs the original pitch (= root key) of the raw sample. Today's software samplers do not need metadata in the file header of the WAV file, but rather take information from the file name of the respective root keys, which are necessary for arranging the samples on the keyboard. In addition to information such as instrument name and type of play, which give you a better overview, it makes sense to integrate the original pitch of each sample directly into the file name.With ten snare samples this may sound like unnecessary hard work, but with sample sets made up of hundreds of individual samples, the hard work of labeling when creating the sample bank in the sampler pays off.


Even if you may never have been particularly sensitive to the subject of noise and background noise with conventional recordings of instruments, sooner or later you will prick up your ears and then let the corners of your mouth drop when you play a non-denoist sample bank. Why is there a constant noise now when I press a lot of keys and then again not when I play a few? The reason for this lies in the addition of the noise when several samples are played at the same time. If you only play in unison - that is, unanimously - everything is smooth. However, if you play polyphonically, with each new voice, not only the instrument sound is added, but also the noise captured in the recording process. The sound of the instrument is hardly noticeable and forms the desired polyphonic sound for our hearing, but the noise of the microphones and preamps used adds up for each voice. Noise, which is barely audible with conventional music recording, quickly adds up to X-fold with extensive sample sets. A piano sample set played with both hands can shoot 50 or 100 voices and more out of the engine with the use of the sustain pedal or long release samples when playing fast. So invest at least a little time in post-processing a sample bank to remove the noise from the recordings. With simple tools or modern denoising plugins, this is not too difficult and even a slight, simple reduction of the noise component by six decibels can make previously problematic, annoying noise fade into the background and save the sample bank.


Even with large, commercially successful music productions, you can occasionally make out a slight background noise if you listen carefully. Since you theoretically only hear a single sound in a three-minute song every three minutes, it might not get on your nerves that much. With samples, however, interfering noises must be absolutely avoided or removed. With a little bad luck, it is quite possible to hear such an interferer every few seconds or even more often. So try to remove every disturbing noise at the latest in the post-pro and be more picky than you were with conventional recordings.

Volume and tuning

In order to be able to playfully control a sample bank, stringent and even sounding raw material is necessary. The more pedantic you pay attention to uniformity, the rounder the bench will sound. If the volume of samples deviates strongly and irregularly from the original, natural sound, you will not have good dynamic control with the finished sample bank. Make sure that there are no major deviations from the neighboring tones within a volume layer. It is just as important that the tuning of each individual note in a sample set is optimal, as you cannot correct an unclean intonation while playing the finished sample bank.

The samples don't just have to be cut ...


Creating a sample bank may seem like a theory-heavy matter, but at the latest when the first specially recorded sound is on a key and thanks to the sampler engine becomes an individual instrument, most of them are hooked and use the sampler and the samples as their very own personal creation Instrument. With the now extensive selection of software samplers, everyone should find suitable software, from free freeware samplers to DAW-integrated sampler tracks. Most products require a little training as a beginner, because the extensive functions of a modern sampler engine cannot be operated quite intuitively. Once you have decided on a particular software, you can find many tutorials and demonstrations on YouTube that make it easier for you to get started and build your own sample bank. If you work with a Mac, you should start with the EXS24 integrated in Logic, and PC users with one of the two programs Kontakt (Native Instruments) or HALion (Steinberg). Their communities have grown quite large over the years, so you can find a lot of help both in forums and on the manufacturers' YouTube channels.