?? Suno.ai AI music, McGurk effect, Synesthesia, and delusions in personal art? ??

Recently, Suno.ai released their version 3 of the text-to-music AI generator. I already had loads of fun with their version 2.? I started a YouTube channel, ‘Every Day as AI Song,’ to turn daily stories into AI-generated music.

The idea was to capture daily stories as AI-generated songs, akin to “1 second every day.” Like life journal but in audio musical form.

Now, Suno released version 3, and it feels like a big step up in quality, especially in terms of the much broader support for genres and songs in other languages. The breadth of things one can do is astounding.? Like this Russian Chastushki song based on the Pulp Fiction script.

Or “Now You Know Your ABC” as a beatbox song:

Or here, beatbox song about startup wisdoms:

You can create acapellas, opera, and even theatrical pieces, beatbox. It’s just hard to control and steer.

But throughout that experimentation, I noticed something very curious.

You can create acapellas, opera, and even theatrical pieces. It’s just hard to control and steer.

But throughout that experimentation, I noticed something very curious.

Creator(initiator? generator? requestor?) versus Listener experience

Some of the songs I created had a profound emotional impact on me. The process of transforming text into music with Suno.ai is captivating. However, when I shared these songs with my wife, they didn’t resonate with her to the same degree. Initially, this puzzled me. I had a hunch that there’s something fascinating happening here.

This mystery began to unravel when I found myself on the other side of the situation. A friend of mine is a writer with a rich family background in poetry. He finds joy in using Suno.ai because it breathes musical life into poems, despite his not being a musician. He relishes the opportunity to hear his and his family’s poetry transformed into music for the first time. And dived in deep.

Occasionally, he creates 60–80 different versions, then blends them into preferred compositions using Audacity before sharing them.

My usual response was to ask for the text first, as it was challenging to gauge Suno.ai’s effectiveness without the original lyrics. I hadn’t thought much about this habit until recently. This week, he shared a new song, and I opted to listen without prior knowledge of the lyrics, noting my impressions and trying to write lyrics out of what I Was really hearing. The result was disconcerting. It was hard to listen, and I got a lot wrong.

This experience raises a question: Is there a distinct difference in how the initiator and an outside listener perceive the AI-generated music? If so, why?

McGurk Effect and Synesthesia

I recalled two fascinating phenomena that illustrate the wonders of our brain’s workings. First, there’s the McGurk effect, a curious illusion where what we see can change what we hear. If someone says “ba” but their mouth moves as if saying “fa,” we might hear “fa” instead, just because of the visual cue. You can experience this phenomenon through various videos online, where the visual and auditory information clash, leading to surprising perceptions. Just try this video while reading different phrase on different loop:

Then, there’s Synesthesia, a condition where sensory experiences blend in unusual ways. Specifically, audio-visual Synesthesia is where sounds might evoke the perception of colors or shapes, hallucinations of sorts. For some, a certain musical note might consistently appear as a specific color, or a voice might trigger a visual texture, creating a rich, intertwined sensory experience that’s unique to each synesthete. I reality all of us have it to some degree. McGurk effect is just one of those that we all share. and it can go in all directions. What we see influences what we hear too.

Does Our Brain’s Wiring Distort Our Experience of AI-Generated Music?

It certainly seems so. The effect that Suno.ai have on me during creation is very powerful. But others do not react the same way on resulting creations. Is it because what I create is personally mine and doesn’t resonate as much with others? Or is it because I know the ideas, texts, and lyrics that were inputs, and the listener does not?

This leads to even more intriguing questions.

How can one confidently edit a song if they already know the?lyrics?

Imagine having a distorted perception of a song’s quality because you’re familiar with the text. It seems like those who edit the song should not know the text beforehand to judge it purely on its own merits. This raises questions about feasibility of solo creators, even with AI assistance. External feedback loops and collaboration still seem irreplaceable.

What does this mean for personal?art?

Art is often viewed as a form of communication. An artist, a master of some medium (be it drawing, writing, singing, sculpting, etc.), communicates important and timely ideas through that medium to others. But with generative AI, something even stranger becomes possible: AI can create personalized versions of art about things that are important to one person. AI serves as an artistic mirror to the deep experiences of an individual. And that art can be incredibly impactful for the individual but might not resonate at all with others. Will this become a common practice?

But if it does, wouldn’t it lose another significant aspect of art? That it’s communal, societal? Art is a medium of communication and unity, bursting our bubbles. Maybe we will have personal art assistants that take societal-level art and personalize it to make it even more impactful for us personally.

Wrapping up

It’s clear that the journey into AI-generated music is deeply personal, and the songs that feel like extensions of ourselves might not always echo as loudly to others.

On top of that, I’ve noticed an intriguing transformation: AI tunes that at the moment of creation held my complete captivation seemed to sound worse as time goes, possibly because the details of their inception?—?texts, lyrics, and the emotions they stirred?—?fade away. As these elements dim in my memory, so does the illusion of song quality.

I‘m keen to learn from your experiences with AI in the arts. Does this story brings any of your “AI and Art” stories to mind? Please share, I am eager to learn?:)

P.S First experiment with posting directly on Linkedin, but its older article from Medium. If you prefer Medium, check out there.