“Creating your own music with generative AI tool”, I wonder what 16 years old me would think about this idea. If someone were to explain to me that I can make pieces of music simply by typing words into computer, I would still have a hard time believing this. It could mean that I no longer had to go to music classes every Tuesday evening, and instead I could write down my music ideas at home and observe what outcomes appear within seconds.
One of the tools that allows such an innovative solution is Mubert. Depending on the desired goal, users can choose different versions of this tool with respective subscription plans. In this blog post however, I will focus only on the basic function called Mubert Render. One of its big advantages is relative ease of use, as there are only two fields that need to be specified. The first field is a text box where users can indicate mood/genre/instrument or even BPM of created music piece. Then, users choose the length of sample and simply click on “generate track”. When describing desired goal just by one word, the results were not that precise unfortunately, as even for key word “acoustic” tracks would sound electronic and would lack distinction of used instruments. Additionally, after changing single words, the results would still sound very similar to each other. However, as sentences grew with more detailed descriptions, created pieces became more interesting and pleasing to the ear. For me, finding a satisfying piece that matched my expectations took a while, and eventually it worked after I specified not only key words, but also the music key for whole sample. As an improvement idea, it would be interesting to add option to divide instruments in given sample and set different key words and rhythms for each of them.
Definitely, tools like Mubert provide an opportunity for creators to expand their music portfolios (Weingrad & Kumar, 2023). It is especially important for smaller artists, as at the start of their career they don’t have enough resources to invest in professional equipment. It can be also useful for people that simply enjoy music and would like to try creating something for themselves for free. Even though AI-created music can’t fully compete with human-created songs yet, it can be used as a source of inspiration and experimentation.
References
Weingrad, H. W., & Kumar, P. K. (2023). Embracing AI’s disruption of the music industry: advantages, challenges and the future. https://www.dglaw.com/wp-content/uploads/2023/08/WLT_Weingrad-Kumar.pdf
You discuss an interesting Gen AI tool that I had not heard of before, so I was immediately interested to read your blog. What I think you did well, is that you were critical of the quality of the results Mubert gave you. By exploring both the possibilities and limitations, I felt like your blog created a comprehensive image of Mubert, rather than creating a sales pitch of how amazing Gen AI is. Given the limitations, I agree that for now, Mubert or similar AI tools can used for inspiration. Whether at some point AI-created music can actually compete with human-created songs, I personally doubt, but who knows.