The rise of AI-generated music has been a topic of discussion among musicians and AI enthusiasts alike. With the launch of Udio, an AI music synthesis service created by ex-DeepMind employees, the fear of automation replacing human creativity in music has become more tangible.
Back in 2002, I ran a music website where visitors could submit song titles and I would write and record a silly song around it. In the liner notes of my first CD release, I mentioned the possibility of computers churning out music automatically, putting me out of business. Little did I know that day was fast approaching.
Udio is a remarkable service that can create high-fidelity musical audio from written prompts, including user-provided lyrics. It covers various genres like country, barbershop quartet, German pop, classical, hard rock, hip hop, and show tunes. It utilizes a large language model to generate lyrics and synthesizes music using an undisclosed method, likely a diffusion model.
However, while Udio’s AI generation capability is impressive from a technical standpoint, it is not without its flaws. The songs created using Udio’s tool may sound half-baked and nightmarish compared to those generated by other AI-powered music generation services like Suno. The high-quality musical samples showcased on Udio’s site are likely the result of significant human input and cherry-picking the best parts out of many generations.
To create a song on Udio, users can enter a text prompt with lyrics, a story direction, and musical genre tags. The AI model then generates two distinct song snippets for users to choose from. Songs can be published for the Udio community, downloaded for sharing on other platforms, or directly shared on social media. Other users can also remix or build upon existing songs.
The launch of Udio has sparked concerns among musicians, as it is the kind of AI-powered music generation service that over 200 musical artists protested against recently. The fear of automation replacing human creativity in music is a valid concern, as many musicians wonder why we need to automate art.
However, AI research aims to replicate art because the imperfect results can still be notable or impressive, which is a characteristic of generative AI. The flashiness and general lack of quantitative rigor make AI-generated art intriguing and captivating. AI has already made advancements in generating still images, videos, and text, and fully composed musical recordings seem to be the next frontier in AI-generated art.
The Udio team has taken measures to ensure that the generated music remains original, blocking tracks that closely resemble the work of specific artists. Although the specific details of the model and training data are undisclosed, it is likely that copyrighted material is present. This raises concerns about the legality and ownership of the AI-generated music.
In conclusion, the rise of AI-generated music is both exciting and unsettling. While the technology opens up new creative possibilities, it also raises questions about the future of human creativity and the role of AI in artistic expression. Whether AI-generated music will ever surpass the creativity and emotional depth of human-created music remains to be seen. In the meantime, musicians and AI researchers will continue to explore the boundaries of this emerging field, pushing the limits of what can be achieved with AI in the realm of art and music.
Source link
