AI music elicits strong opinions on both sides of the debate. There is the argument that much of today’s music is already heavily influenced by effects, samples, and enhancements, making it essentially AI-generated without explicitly being labeled as such. A lot of contemporary music features heavy synthetic beats and autotuned vocals that sound nothing like the original singer’s voice. With all the effort that goes into producing a “perfect” song, is AI really that different?
A classic example of this is humorously illustrated in a certain video, which, though made as a comedy skit, reflects some truth.
On the other side of the argument, the perception is that AI music involves simply prompting an AI to “make a hit song about cowboys” and then publishing it. While someone could indeed do that, it is unlikely to result in an actual hit song. This approach is often seen as the lazy way to make AI music, and many people are doing it.
AI Music Soulless and Unemotional
Critics argue that AI music lacks emotion and will flood the market with generic, soulless music. They claim that it is not truly creating music, and they have a point. However, not everyone uses AI music in that way. There are various methods people employ, such as:
– Writing lyrics and using AI to generate the music and vocals
– Writing lyrics and getting AI to generate only the music
– Writing lyrics and getting AI to generate only the vocals
– Uploading a rhythm or short music piece and getting AI to generate lyrics for it
– Uploading a rhythm or short music piece and getting AI to complete and produce it
– Uploading a vocal track and getting AI to add music to it
– Uploading a vocal track and getting AI to change the gender or sound of it
– Creating partial lyrics and getting AI to complete them
– Any of the above and then producing them in a Daw making adjustments
I could elaborate further as there are many more variables to consider. The point is, not everyone simply inputs a prompt and receives a satisfactory result. Even when they do, it is rare that the initial output meets their expectations. For some, this means experimenting with various elements such as modifying the prompts, altering lyrics, or incorporating musical instructions. Examples include adding guitar solos, rising vocals, short instrumentals, or for those more musically knowledgeable, even specifying chords to use.
AI Music Apps
Suno
As I discovered AI music applications and websites, I experimented with lyrics I had written many years ago. I have a folder with all of them written down. Some lyrics managed to sound close to the sound I had envisioned in my head, bringing my song to life. I had not looked at the lyrics in years. Not all of them worked as I wanted, and others took dozens of attempts to get them sounding close to what I wanted or at least be happy with the sound.
I started with Suno and can say that the speed at which it generates songs is amazing. If you provide good instructions with the lyrics and style, it can produce some fantastic results. Sure, some of them sounded ordinary, but overall, with a bit of experimentation, I was getting good results. On occasion I managed to get more than one version where I couldn’t decide which was best as I liked them both.
Udio
The next one was Udio. I really struggled with this one. The instruments and vocals sounded clearer, but I couldn’t seem to get the sound I wanted from it with my lyrics. Another problem was that it only generates up to 2:11 as opposed to Suno’s 4 minutes. Both have the option to extend. One interesting thing about Udio is that you can earn credits by listening to other songs on the platform.
Mureka and Riffusion
Another AI platform I tried is Mureka, although I haven’t done much with that one lately but it was promising. A new one I recently discovered is called Riffusion. It is a bit like when I first found Suno, as I am liking a lot of the songs I’m generating with it. It is hard after already generating songs for my old lyrics to get better, but now I’m experimenting with new songs, mostly AI then adjusting them
Jazzabelle Rox On Streaming Platforms
After dedicating significant time to AI music, I invited Jazzabelle to try it out. She loved it and wanted to create songs for her social media and streaming. Her musical style differed greatly from mine. She would come up with partial lyrics, use chat gpt to fill in the blanks, and then use Suno to create the song. From there, I produced it. I found lyrics in some songs that I modified before recreating them in Suno before producing with Bandlab. I suppose that is what producers do to achieve the best sound.
I then distributed the songs to streaming platforms. All of her songs are under the artist Jazzabelle Rox. Interestingly, I have not distributed most of my own songs as I am trying to perfect them. I have some songs with female vocals, country, alternative rock, pop, etc., and I plan to have different artists for each genre. It is challenging to decide what to call them. Currently, I only have one artist so far, The NMC. I have dozens of songs that I just haven’t got ‘right’ yet.
Conclusion
AI music is here to stay and it will always have it’s haters. Is that any different to auto-tune, or when synthesizers first came into music? What I can say is that as it is only in early stages it is sure to get better and better. There is already many great AI songs out there and a lot of what many would call rubbish. Is that any different to how it has always been with conventional music? I think not. This may make it easy for anyone to come up with a sound that becomes the next big hit, but how will it be perceived? What do you think?
Be First to Comment