It all started with a post that happened to float down my social media timeline. The song, which someone had apparently created using AI, sounded so natural, like a new release from a real artist, that it immediately caught my ear. "Can an individual really make something like this?" Unable to resist my curiosity, I immediately started playing around with a tool that was generating a lot of buzz: Suno.

I typed a few short words like "cheerful song" or "melancholic melody" into the prompt field and hit the generate button. In just a few seconds, a melody rose up, accompaniment kicked in, and even vocals began to flow. At first, I was simply thrilled by the sheer speed at which sound appeared before me.

That was where my exploration truly began. I tried Udio, renowned for its expressive vocals; experimented with cutting-edge technology like Lyria 3; and even ventured into platforms like Mureka and Somio, which integrate everything from music creation to music video production. Every time I switched tools or mechanics, a fresh sonic texture was born. The more I experimented, the more new surprises and discoveries I stumbled upon.

From then on, my days of repeatedly hitting the generate button began, almost like pulling for characters in a gacha game. "If I press it one more time, what kind of incredible track will pop out?" What I felt back then was a pure sense of anticipation toward the unknown, brought by sheer chance.

A Growing Sense of Disconnect: Emptiness in a Sea of Randomness

However, as the number of generated songs grew from dozens to hundreds, a subtle sense of disconnect began to take root in my mind.

"I feel like I've heard this somewhere before."

The audio fidelity was polished, and the song progressions were smooth. Yet, every track began to feel like it shared a similar structure and a similar atmosphere. What lay before me was simply a lineup of "averagely pleasant sounds" that the machine had learned everyone tends to enjoy.

What felt most lonely was that no matter how high the completion quality of a song was, it somehow felt completely detached from me. I was the one who pressed the button, but I wasn't the one who chose the sound. Leaving everything to chance yielded not a sense of accomplishment, but a hollow emptiness.

Having an Intent: Pulling Sounds Closer with Words

"Let's stop leaving this to luck." Having made up my mind, I decided to change my approach before hitting that generate button.

Instead of just writing "a sad song," I started feeding it more specific and constrained descriptions.

• Focus on a slightly raspy, clean electric guitar instead of a piano

• Cut the bustling drums to create a quietness reminiscent of a rainy midnight

• Use a warm, human singing voice reminiscent of old-school rock

When I did, the sounds coming back from the other side of the screen began to change drastically.

There were moments when combining unexpected instruments sparked a miraculous harmony. At the same time, I learned that feeding it contradictory words resulted in muddy sounds and chaotic, bizarre tracks.

The feeling of my own words directly reflecting in the sound—including the failures—brought a genuine sense of "dialogue" that was completely absent when I was just playing the lottery.

Anatomizing My Own "Preferences" Before Putting Them Into Words

The more detailed I tried to make my instructions, the harder I hit a specific wall. It was this: if you don't know what kind of sound you actually want to hear, the words simply won't come out.

Refining words wasn't an exercise in writing a command spec sheet. It was the very process of dissecting the vague "likes" and "images" inside myself and giving them form through language.

• Why does this song feel unsatisfactory? (Is the bass resonance too light?)

• Is the "sadness" I'm seeking about a slower tempo, or does it live in the negative space between notes?

AI wasn't a tool to spare me the effort of thinking. Rather, it stood like a mirror, sharply questioning my sensibilities and choices: "What do you really want to do?"

Beyond the Gacha: The Human Work That Remains

The charm of AI music generation lies in how easily anyone can produce sound. Yet, bringing a small piece of intention into that overwhelming convenience—refusing to leave everything to luck—is where I feel the true joy of creating resides.

Picking a miraculous track out of a sea of randomness is certainly one way to enjoy it. However, taking the faint hint of a sound inside your head and gradually pulling it into reality through the power of language...

What remains after pressing the generate button is not just a filtering process. It is understanding what moves your heart and unyieldingly choosing your words. I believe that is precisely the precious work left for humans in an era where we make music alongside new tools.