No ratings yet.

AI has certainly brought about a wave of intrigue within the music industry, owing to its remarkable capabilities, particularly in music production. However, it also presents limitations. My conviction in this matter arises from an experiment I conducted with a friend who has a livelihood as a music producer. Given my unfamiliarity with music production, I believed that collaborating with a music producer would provide more value to the experiment.

Traditional music producers possess a unique blend of creativity and technical expertise, including keen listening skills, a profound grasp of recording technologies, and extensive musical knowledge (Malta, 2021). As we started our short journey to find suitable AI tools, an immediate limitation became apparent. No single AI could seamlessly integrate the numerous steps required for music production, including melody creation, percussion arrangement, lyrics composition, and vocal creation. Consequently, we had to aim for distinct AIs to fulfill each of these tasks. We ultimately enlisted the help of four AIs: AIVA, The Infinite Drum Machine, AI Songwriter, and Uberduck. These AI tools gave us a large volume of possibilities, based upon the prompts we provided to guide them towards our desired outcomes. For instance, we used prompts such as “create vocals in the style of the famous rapper Drake” and “generate lyrics based on our childhood” to instruct the AIs effectively.

The end result of our experiment was undeniably impressive, but it came with a caution. The AIs, with their current online resources, were not self-sufficient in producing music. My friend and I found ourselves making numerous adjustments at each step of the process to fine-tune the AI-generated music to our satisfaction and to reach an acceptable result. I am excited to discover what will be the future outcomes of AI in music production as it is for now at the early stage its development.

Sources:
Malta, S. (2021). Retrieved from: https://www.edmprod.com/what-is-music-production/

No ratings yet.

This AI experiment originates from one of my favorite hobbies: DJing. I consider myself an amateur DJ, mainly performing “shows” in front of my friends. However, I now want to perform real shows in nightclubs. Therefore, I’ve started to think like a DJ, to make my DJ sets more enjoyable. The primary role of a DJ is to create innovative and captivating music that generates and sustains emotions on the dance floor. This entails extensive music research to discover new and lesser-known songs available on music platforms such as Spotify, Beatport, Soundcloud, and others. 

I began following these steps and decided to use generative AI, specifically ChatGPT, to see if I could make music investigation less time-consuming and, more importantly, improve its quality. I used ChatGPT in two specific ways: First, I employed ChatGPT to help me find new and lesser-known music based on precise criteria. For instance, I specified the music style (House, Electro, Groove, etc.) and provided some music examples to support the research. I also indicated whether vocals were present and that the songs had less than fifty thousand views on music platforms. Second, I used ChatGPT to assist in preparing my DJ set by requesting a logical and coherent sequence of music based on a list of fifty songs and their corresponding beats per minute (bpm).

Once ChatGPT fulfilled all my requests, I used the suggested songs and followed the provided order during a DJ set to assess whether it could adequately improve my future shows. Based on the preparation process and the DJ set experience I concluded that: First, ChatGPT saves time in preparing a show and suggests songs matching my tastes. However, there is still work to be done to find other lesser-known music and ensure that the DJ set matches the emotions of the audience. In fact, ChatGPT cannot completely feel my emotions in terms of music tastes as well as of the audience, which requires adaptation according to the atmosphere.

I am interested in hearing other music fans’ thoughts about my experiment and whether they also used ChatGPT in a similar way.