Recently, I've been experimenting with Suno and Udio, two generative AI tools that create music from prompts—like Midjourney for images. Surprisingly, the results are quite good! Suno and Udio each have slightly different interfaces and interpret prompts in their own ways. I've noticed that Suno tends to deliver more polished results, albeit the most predictable ones. Udio, on the other hand, is less polished but occasionally produces interesting, unconventional outputs.
Music creation traditionally required decades of dedication, and complex collaborations—from composers to musicians to sound engineers—to produce a masterpiece. Now, an AI can generate a decent song complete with lyrics, prosody, instrumentation and mixing in about 30 seconds. This means anyone can create music without years of training or the need for extensive manpower. The market is about to be flooded, big time! 😅
Some people have expressed concerns about how this will affect platforms like Spotify, but I believe that in the long run, most people will not need a “repository” of music when you can endlessly generate new song in any style on demand.
So, does AI really make good music? No, it produces acceptable songs. They will get better soon enough though.
Will it create truly great music in the future? Probably not. As the technology improves, the output will become cleaner and more refined, hence bland and forgettable: AI is optimized to search for, and output the most likely strings of words, chords, or pixels. But, the best music often embraces the unlikely: the weird stuff, the accidents, the dissonant, the noise and the silence.
AI may soon replace acts like AKB 48 and BTS (thank God!!), but I am curious to see if it will ever replace Frank Zappa. 😬
The day it does, I'll buy a ticket for the Coast.
