CHOR JAVON TIMING AND MIXING
Work example about composition of computer/digital music: Chor Javon timing and mixing.
When making music in a computer we have the possibility to integrate parts of an existing music theme (in the form of an audio clip) into
our own compositions. Of course, this is probably a bad idea if we are creating a professional composition for commercial use, but we should
have no problem as long as our composition is intended only for personal use. To illustrate this example I have chosen an audio clip that I
once found in an online music store; it belongs to one of the countless versions made in Asian countries of a Tajik song called
Chor Javon. To begin I have to present the audio clip in question:
chor_javon_timing_mixing_original_sample.mp3
The melody of Chor Javon is based in a pentatonic scale, as it has been typical in traditional music. It is not surprising that when working with these short scales a composer may try to compensate the limited choice of notes with repeating hits in a same note. During the largest part of the song only three notes are played (C, D and E) and a same key is played up to five consecutive times; then, notes Bb and A are used to close the composition at the end of the second melodic phrase (Chor Javon includes two different melodic phrases but the sample presented here includes only the second one, sung twice). Despite of its simplicity, the melody achieves expressiveness thanks to the timing used when playing the notes.
However, this tutorial will not deal much with melodic rhythm, as its main purpose is the synchronization of drum beats. For completing this task succesfully it is essential to know the tempo which was used to create the reference audio clip, so we would be able to build our own drum tracks. Also, if we know the scale used in the composition, we could attempt to include additional melodies. Besides what is shown in this tutorial, I created my own version of Chor Javon even if the score was not readily available in the internet. Since I knew of a video where someone plays the melody in a keyboard, I used this video and a slow-motion effect to write the score in a piano roll; the most difficult part was to draw the notes in their proper position, because otherwise the personality of the melody would be lost.
In the version arranged by myself I used some creative licenses as I thought convenient. I supposed that the cell structure of the piano roll was not flexible enough to represent the timing of the original composition, but still I preferred to snap notes to cell to make things simpler to understand. I also introduced some note hits that did not exist in the original composition; in the first and third measures there were only four note hits, but I used five because the rest of measures used five as well. The result was a melody that, even if in a purist sense might not be Chor Javon, is easily recognizable as Chor Javon.
Now it is time to focus in the rhythmic foundation of the reference audio clip. The following pictures show how one of the measures of the audio clip looks like, either in waveform or spectral view, in the display of Cool Edit Pro 2. There is a setting in this audio editor for switching the horizontal ruler of the display between minute/second and bar/beat format. In Cool Edit Pro 2 this can be done by right clicking on the ruler and then clicking on Display Time Format to finally select Bars and Beats. The large numerals located beneath the ruler represent BAR:BEAT.TICK; this means that, for example, 2:2.08 indicates that the current position is the 8th tick of the second beat of the second bar. In this example, each bar has four beats and each beat has 16 ticks, but the grid in the display will vary according to the level of zoom.
In the left end of the ruler there is indication of the tempo, which in this case I have adjusted to 134 beats per minute, value which I had to find by simple trial and error procedure. Since modern danceable music usually has values of tempo ranging from 120 to 150 beats per minute the search is not so difficult; we just have to try different values until seeing that all the drum hits in the audio clip, displayed either in waveform or spectral view, match as best as possible the vertical lines on the grid. As it can be seen in the pictures, the coincidence is practically perfect. By switching the numerals to minute/second format we can see that the duration of each measure is 1 second plus 791 thousandths, as seen in the lowermost picture.
At first, determining the tempo seemed good enough to fulfill the task, but I was intrigued by the fact that the drum hits did not have a fully uniform pattern. It becomes obvious when looking at the spectral graphical representation that every beat is divided into two parts by the drum hits, and hence each measure should have eight drum hits, but in some beats there is an additional hit. If I wanted to recreate the pattern as faithfully as possible this detail should not be overlooked. After careful examination I found four different patterns which repeated consecutively in the same order, as seen in the following picture.
Once we know how the different patterns are arranged we can draw them in a piano roll. Regarding the source audio clip, it is imperative to define with precision where its starting point should be. To achieve a good synchronization between the drum hits on the audio clip and the drum hits on the piano roll, we have to trim, with the maximum precision possible, the beginning of the audio clip to match the beginning of the measure that is nearest to the starting point. Finally, we have to properly snap all the notes in the grid of the piano roll, either those used to trigger the drum hits or those used to trigger the audio clip.
Then we have the choice of which sound patches to use when creating our drum track. It is quite unlikely to be successful on properly isolating the original sounds used in the source audio clip to feed a sampler with them, so using the sound patches provided by a drum machine should be the most common approach. If we want to recreate the drum track with precision, we can choose sound patches that are similar to those used in the source audio clip and place them in the matching positions; otherwise, if we do not want to invest more time on this, we can use any sound patch of our choice that is able to effectively mask the percussion hits of the source audio clip.
For creating the drum track used in this example I took samples from the source audio clip, even if there was no time throughout it where percussion hits were properly isolated; they were merged with voices or other background sounds but I felt that this created an interesting effect. To synchronize the percussion hits with those present in the source audio clip I set the tempo of my FL Studio project to 134; in the picture above we can see that when positioning the marker in the line that divides the first and second measures the time display shows 0:01:79 (1 second plus 790 thousandth), which is very close to the value shown by Coold Edit Pro. Thus, the timing work was succesfully finished.
But regardless of this, I was not happy with the drum hits present in the source audio clip. They were numerous and complex because of the four different patterns and I felt that they together with my own drum hits would fill the sonic space in excess. Cleaning the source audio clip was impossible and therefore I tried to recreate the patterns in the piano roll. The result was far from satisfactory, so in the end I just created a drum track populated with numerous hits, which should be able to dissimulate the percussions of the source audio clip, specially when lowering the volume level of this one.
Finally, I found that my own Chor Javon melody matched quite well the structure of the chants in the source audio clip, so I added this to the mix, along with a bass track whose note patterns are based in those used in the main melody. Eventually I found that the sound of the source audio clip had a muddy quality, so I used an harmonic generator to add brightness.
chor_javon_timing_mixing_final_mix.mp3
Wave slicing
Wave slicing is the process of taking audio samples from a longer sound clip. Simple wave slicing is effectuated in sound clips where only a single instrument plays (a cappella tracks); this can be done either through manual action or by means of specific software which may automatically detect drum beats or musical notes standing on a silent background. Complex wave slicing is effectuated in sound clips where the different instruments and voices are mixed together, which renders this task difficult, if not impossible, to be fulfilled properly or with complete success; in this case, the samples can be taken through manual action only.
When attempting complex wave slicing on a music theme of our choice, the first task required is to find its tempo; only after being sure about the correct estimation of this parameter should we begin to take any samples. To find its tempo, the music theme has to be loaded into an audio editor which can display bar and beat marks for any tempo value chosen; the task consists of adjusting this value until a hit that we take as reference occupies the same position on every bar, without any gap from bar to bar. Then, we can adjust the position of the reference hit within each bar by just adding or removing small stretches of silence at the beginning of the audio clip.
Once the aforementioned task is done with precision, we can begin to take samples by using as reference the bar and beat marks. It is expected that when these samples are used in the pattern arranger of any music virtual studio, where they should be adjusted either by bar or by beat, the connections between them will be smooth and their rhythm will be on synchrony. The samples should be stored in WAV format because this format ensures the preservation of quality after multiple editions, and also because compressed formats might add silence at the beginning of the file, altering so the necessary synchrony.
It often happens that the source audio clip has noise in the background; by "noise" I am not only referring to actual noise, but also to any subtle ambient sounds that the producer might have added to the composition. The problem caused by noise is that in the transitions between "noisy" samples and areas of clean background a more or less sharp contrast will be perceived. To clean the noise from a sample, we can use a graphic equalizer to attenuate the higher and lower frequencies, where noise usually dwells. We can also use a noise removal processor to help with this task. During these processes we have to find a balance between the reduction of noise and the perceptible loss of quality that the remaining signal will suffer.
A kind of task that sooner or later we would want to perform is the separation of a drum track from other musical sounds. There are not many options regarding the techiques available to achieve this, and the results usually range from unsatisfactory to null. In such cases, wave slicing offers an alternative. The following picture shows a drum pattern which has been built by means of this technique, to resemble a drum pattern which was mixed with vocals in an MP3 sample file. The pattern is variable and repeats itself every four bars. The purpose of this meticulous work is to achieve a drum pattern which allows to integrate the sample vocals into any custom musical composition (as long as the sample is properly trimmed to achieve rhythmic synchrony). When synchrony is good enough, the drum pattern can be played simultaneously with the audio sample that it tries to mimic, albeit this might cause the drum beats to sound too bold.
The procedure begins by making a copy (A) of the sample vocals and removing from it anything except the drum beats; for doing so we select the areas to be removed and replace them with silence. Once the drum beats are isolated, we can make a copy (B) of this file to take samples from it. Then, in the A copy we have to decrease the volume level to the extent that the drum beats look very small in the waveform display, so they become hardly audible but still visible. From the B copy we can select a sample of each kind (bass drum, stick, hi-hat...), ideally one in which foreign sounds are the least prominent, and by using an equalizer we can remove the unwanted frequencies, so the sample sounds clean. Once we have clean samples, we can use them to rebuild the drum pattern in the A copy, by pasting - in mix mode - the samples on the corresponding places. If the transitions between silences and beats sound too sharp, we can apply some kind of delay effect in the track to try to smoothen the sound.
Finally, I will show a method for synchronizing a melody arranged on a piano roll with a melody that is sung in an audio sample; this process is assisted by the spectral view that is available in a professional audio editor, as we can see in the following picture. If the audio sample is not excessively busy with disturbing sounds, we should be able to see the position of individual musical notes; this way we can arrange musical notes in the piano roll whose timing will match perfectly that of the notes in the audio sample.
Going further with this technique, we might as well slice individual notes from the audio sample to try to create our own combinations; but given that these notes are sung by human voice, to achieve a natural result it is required that the intonation - and not only the pitch - of each note matches well its position within the new melody.
Note: this example does not correspond to the music theme depicted in the previous example.
This video tutorial explains in a more visual way how to slice waveform samples manually from a music track.