In 2020, hip-hop is the most prominent genre in the United States and, statistically speaking, the most popular. Over the past nearly half century it has evolved from a local New York City phenomenon to a ubiquitous organism of music and culture that has been called “America’s musical export” to the world. While modern hip-hop has swiftly evolved and grown far from its roots, the history still impacts today’s sound and culture and has influenced numerous other genres along the way both in the United States and internationally. While countless factors built the early microcosm of hip-hop, one ingredient was arguably the most vital – even long before it had a common name – and that was the “break”. 

            Breaks can be referred to interchangeably as “breakbeats, break beats or breakdowns,” but all these terms essentially refer to the same fundamental aspect of hip-hop and its production. A break is an instrumental or percussive section of a song traditionally derived from or related to down-time – being a “break” from the main parts of the song or piece. It is generally the part of a track with minimal instrumentation and no (or minimal) lyrics. A breakbeat refers to the sampling or reuse of breaks as drum loops and their subsequent use as the rhythmic basis for hip-hop and rap. 

            For many, sampling has shifted from analog (vinyl) to digital, opening the door for vast amounts of music and audio across the internet. While this art of sampling has evolved from digging through record crates to digging through unfathomable bits of musical data, the listener, DJ, or the producer still has to sift through hours and hours of music sometimes to find a suitable sample or break. However, with the advent of virtual intelligence systems, the methods of sifting can continue to evolve and maximize in efficiency. I have proposed and developed a neural network for this task to quickly and systematically sort through immense libraries of music for the purpose of automatically classifying suitable portions of songs that fit the classification of breaks. Paired with an additional utilization tool, this network facilitates extracting samples and creating hip-hop beats more easily and rapidly than ever before. 

            In doing so, a small library of breaks and their respective original songs was utilized to train a deep learning network to analyze the spectral images of these given audio clips with respect to frequency over time. This modified convolutional neural network (CNN), named the Perceptron Audio Classifier (PAC), returned a recognition and proper differentiation rate of 98.89% with loss factors as low as 2.49%. The results of this CNN were then used as the basis for what I call the Break Identification Genius (BIG), a utilization tool designed to analyze entire songs by systematically segmenting them into bars according to a BPM (beats per minute) analysis and an onset detection algorithm using Google Colab. Essentially, entire songs were segmented, and the PAC model was utilized by BIG to determine whether every single section was considered a suitable break or not. Once a suitable break was found, these bars of the songs could be extracted and easily looped to create a new beat for hip-hop. Together these methods are the basis for the Break Boss concept, an automated breakbeat finder and inspirational tool.