Neural networks are not programmed, they are trained. They learn by themselves. During training, they read the data (hundreds of lyrics manuscripts from Sabotage) again and again, over and over, in different ways, and adjust, trying to distil the content they are meant to learn. That is exactly what was done with the manuscripts. The work began with a Language Model based on Recurrent Neural Network, which deeply looked at all material left by Sabotage and, from that, it created new sentences that were validated by his family and partners. Thus, the lyrics of "Neural" came together, and it got its melody from RZO, the first rap in the world created by an artificial intelligence. The song was released and promoted by Spotify as part of its portfolio.

We developed and trained a neural network using the intellectual production of Sabotage when he was alive, thousands of manuscripts and lyrics. After that the machine started to learn and produce new, original, ideas, new rhymes. The machine began to express itself as Sabotage and, little by little, got to an advanced level of accuracy, as if the artist was relearning how to speak. A group of close friends and family helped the machine to a better understanding of the artist's style by giving realtime feedback to validated sentences. In the end of this process the lyrics of "Neural" were ready, later recorded by RZO.

From seventh to the first most listened genre to on the platform. We definitely made users discover beyond the new rap and hip-hop releases in Brazil. We increased by more than 470% Sabotage's consumption, the main artist of the campaign. We still generated more than € 484k in earned media, doubling the amount invested in paid, and increased by 22% Spotify consideration rate among users impacted.