Beat DNA: A Rhythmically Searchable Loop Database

For literally centuries, the metronome has ruled the world of music with the iron-fisted persistence of time itself; unyielding, constant and when its passing is given the fullest of attention, well…boring. This is not to suggest that strict adherence to tempo is a bad thing, quite to the contrary. Keeping within the confines of a song’s timing has kept the world’s greatest musicians from obscuring or deviating from the musical intentions of composers the likes of Tchaikovsky, Mozart and (perhaps, especially) Bach.

The very predictability of repetition is actually coded into our brains, fortifying and creating connections that assist in communications amongst the different areas of the brain, improving our memory recall and learning ability (which is one of the very things that makes music in schools so important, but I digress).

An even finer point can be made with regard to groove or “feel.” A primal function within the brain is to subconsciously assure itself with repetition, freeing up the mind for other tasks. A steadily ticking clock, for example, can be tuned out after a bit if you have other things you’re doing. However, if a tick occurs slightly ahead of the predicted time it will subliminally draw our attention to it, because it arrived earlier than expected. This results in an almost imperceptibly anxious feeling. In music, this is known as “drive” or playing “on top of the beat.” Conversely, should a tick occur after our brain anticipated its appearance there is a microsecond of uneasiness immediately followed by a feeling of relief, kind of like that moment where you suddenly wonder “Where are my keys?” and then realize they’re in your hand. We tend to remember that rewarding feeling of relief, resulting in a relaxed sense of well-being.  This is often referred to as playing “back” or behind the beat.

Metronomes are ticking clocks. So are click tracks. The human heart is not. Grooves aren’t either.

And, yet, with the exception of classical and jazz genres, the majority of releases in nearly every form of currently popular music from hip hop to EDM, pop, rock, and even country or folk are recorded to a click and laid out on a strictly adhered-to grid. Most of this is due to convenience and necessity, especially in this time of social distancing. Some of us can simultaneously lock to a click and impart the groove effortlessly. Others need a drummer’s feel to sync with. In the absence of a live drummer at the outset of a song’s recording, the best solution available is drum loops.

Drum loops can be a simple drum machine pattern (which is often still too rigid a feel) or a recorded sample of an actual human drummer playing a groove. These can vary in length from one bar to an entire song form. For the purposes of this article, let’s use examples of eight beats (two measures) in length.

Pictured here are three iconic drum loops that were all performed by real drummers. (Figure 1)

The first is from the intro to Michael Jackson’s “Billie Jean,” performed by Leon “Ndugu” Chancler. Upon first glance it seems to be a very simple beat, with the kick drum on the 1^st and 3^rd beats and the snare on the 2 and 4. The closed high-hat plays 8^th notes throughout. I say seems, because it starts out right on the click, but soon falls behind the beat, only to catch up again as loop repeats. This creates of a constant feeling of tension and release.

The second example is the intro to Aerosmith’s “Walk This Way” performed by Joey Kramer and sampled to multi-platinum effect by producer Rick Rubin for rap artists Run-DMC. While only slightly busier than “Billie Jean,” it’s notable for its unusual placement of an open high-hat on the first beat, which drives the loop forward with an explosive hiss.

Finally, one of the most famously used breakbeats ever is James Brown’s “Funky Drummer (Pt. 1 & 2),” performed by the legendary Clyde Stubblefield. Compared to the previous examples, there is more happening between the downbeats than on them, with its offbeat kick drum and high hat patterns bouncing off the snare’s ghost notes, resulting in one of funk’s most unique and defining rhythms.

Three grooves with distinctive musical personalities. Representative of an infinitesimally tiny fraction of the rhythmic possibilities we can avail ourselves of as artists, writers, DJs and producers. As the desire to utilize “real” drum performances in our recordings has grown, so has an entire subset of merchandise within the music retail industry’s canon of products: Loop Libraries.

What started as a group of musical archeologists earnestly digging through crates at record stores in an effort to rediscover and repurpose these snippets of musical magic became big business, as people realized that even the coolest beat repeated over and over can often sour into stagnation. Drum performances need to introduce variation in order to evolve with the song over the course of its duration. Intros, verses, pre-choruses, choruses, bridges, solos, and vamps are integral building blocks of a song’s structure. These sections often need further deviations within them, to break free of the potential monotony of their inherent stasis. Crash and ride cymbals, tom fills, and subtly altered versions of the groove are all a part of imparting a realistic and musically satisfying result.

I have roughly 60,000 drum loops on my computer, equaling about 80 Gigabytes of data. This means when it comes to options, be they stylistic, tonal, meter or groove oriented…I’ve got more than plenty.

So here I am, feeling creative and inspired. I’m ready to choose a beat that I can build upon with additional or alternate parts later. For now, I just want a feel to play with and lay down the song with my guitar and vocal. I know the basic pattern, tempo and style I want. So, where do I start?

“Ay,” as Shakespeare’s Hamlet so aptly pointed out, “there’s the rub!”

I open the Loops folder on my Windows desktop where all of my drum and instrument loop titles reside. I open the LoopLoft folder where I have 51 gigs of loops from 91 different collections purchased from www.thelooploft.com and start looking for something close. I select NateSmithV1Drums_WAV. I go into its Audio Loops section and find something that looks promising: a folder entitled “PurePop_112bpm”. I go into it and select the first loop listed, “Chorus01_PurePop_112bpm” and preview it. Pretty cool, but nowhere near the beat I’m after. I look at another iteration called “Verse01_PurePop_112bpm” just in case, but although the tone of the drum kit is perfect for me, this beat is not the one. I back out of that folder and look in another called “BigBass_90bpm”, listening to a loop named “Verse01_BigBass_90bpm.” Nope. After further exploration, I eventually back out of that title and into one labelled JoeyWaronkerVol3_WAV, go into Audio Loops and find a folder called “Thrifty.” I preview “Thrifty_Verse01_115bpm” and I really like it and the freshly distinctive sound to it is a bonus. I try playing the song to it, but it just doesn’t work well with it, so I back out again. For two more hours I hunt and peck, spiraling ever downward into the rabbit hole until I look up at the clock and realize I’ve got somewhere else I have to be. Frustrated, I power down my computer with the feeling that I just wasted all of that day’s allotted creative time with absolutely nothing to show for it!

Sound familiar?

To be fair, I believe the thinking on the part of loop providers has always been that in listening through the loops, one might eventually be stricken with an inspiration, no doubt a holdover from the “crate digging” paradigm that loop collections evolved from. While that creative model can and does pay huge dividends when it happens, it is still taking the long way around the barn when it comes to carts and horses.

Which brings this discussion to an idea that I’ve held onto for the last five years or so: A rhythmically searchable database for loops that reside on one’s computer. I like to refer to it as Beat DNA.

Why have I been sitting on it, you may ask? Because this whole time I have believed the arrival of this product was imminent. Look at all of the technologies that analyze and even manipulate audio:

ReCycle! by Steinberg and Propellerhead detects and isolates all of the hits in a loop and chops them into individual segments, called “slices.” In addition to giving the loop the ability to be sped up or slowed down without affecting its pitch, slices can then be moved forward or backward in time to adjust the feel, or even placed into a different order, to create variations on the beat. The slice “markers” can even be exported to a MIDI file, which we will touch on later.

ProTools’ Beat Detective cuts and nudges drum hits toward the grid.

Studio One by PreSonus has long had Audio Bend markers that detect hit transients for subsequent Groove Quantizing, allowing users to snap their own custom grid based on the feel of another instrument’s performance.

Melodyne from Celemony can also detect and manipulate the timing of rhythmic elements, as well as convert and export audio into MIDI format. Their latest version (5) can also now detect and manipulate sibilance, which could mean it’s not far from learning the difference between high-hats and snare hits, for instance.

There are dozens more examples, but hopefully you’re smelling what I’m cooking here.

In theory, this is how it you’d set it up: on the established user side (people like me who have a library of loops already), Beat DNA would be purchased and installed onto the computer where the loops reside. Once instantiated, the program would locate and analyze each loop, placing markers at each hit and identifying its source (kick, snare, high-hat, crash, ride, etc.), this info is then tagged and cataloged in a number of ways, so that a search query would be able to identify loops by where these hit occur, what instrument the hit belongs to and its proximity to the beat or subdivision of it the hit lands on. Yes, I do think that this would take a long time the first time you use it. However, the actual data one would have to subsequently troll through once their collection of loops was logged would be tiny little packets of markers, their positions and other identifiers which even a smart phone could power through quickly. Once standardized as a format, this encoding could already be added on the manufacturing side, so the markers would already be embedded into the loops. Much as Sonic Foundry did with “Acidized” files in the past.

Figure 2

Figure 3

Figure 4

Also, in theory, this is how it would work: let’s start easy and say we’re looking for something like the “Billie Jean” beat from our example. The Beat DNA search engine would have an easy-to-use drum pattern step sequencing map. In this case you would start by placing Bass Drum hits on the 1 and 3, Snare on 2 and 4 and 8^th notes all the way across on the high hat (Figure 2). “Walk This Way” differs only slightly (Figure 3), But “Funky Drummer (Pts. 1 & 2) will clearly take a bit more effort. (Figure 4). Once the pattern is entered, other factors can be edited to restrict or increase the results, based upon the likelihood of other hits included or blocked and a hit’s proximity to the beat in front or behind by percentages, i.e. searching for a “Billie Jean”-like pattern with 70% probability might return a few dozen to a hundred examples in querying my personal collection, but I’d be able to go right to them and assess their usefulness toward my purposes without blindly grinding through thousands of candidates. This could save endless hours of my time (which incidentally, can be expensive for clients) and free up resources that could be better applied toward more creative pursuits.

This could also be a boon on the retail side, as customers could visit a website that hosts a Beat DNA search engine and find exactly what they’re looking for and purchase the entire library or just a few loops a la carte, opening up myriad new revenue streams in the process.

Would it be easy to do this? Maybe, maybe not.

In an abstract titled Indexing And Querying Drum Loops Databases from 2003 (https://perso.telecom-paristech.fr/grichard/Publications/Cbmi.pdf), authors Olivier Gillet and Gael Richard propose the possibility of verbally “beatboxing” your queries to find loops within your collection, but seem to get bogged down by syntax and its translation for the purposes of audio searches. I think that is a great idea and it’s exactly where my wish list first started, but that functionality might need to stand on the shoulders of a simpler design. That is why I keep going back to programs that are able to export to MIDI, resulting in smaller, simpler files extracted from the cataloged audio’s markers which can subsequently be traversed in an almost skeletal way, even utilizing the general music standard for instant cursory playback, providing users a glimpse at what the essential groove is before committing the system resources required to process digital audio. This downsizing of data could also facilitate cloud-based technologies as well.

Bottom line: this is a million dollar idea, though I doubt I’ll be the one receiving a million dollars for it.

But I need this tool now and so do thousands of other content creators, music-based and otherwise.

Sven-Erik Seaholm is an award-winning independent producer, singer, songwriter, and recording artist. Please share with him your thoughts and expertise on this concept. kitschandsync@hotmail.com