[Vocals and First-Year Students] 18. The Secrets of Songs Revealed by AI

*This content was translated by AI.

Recently, many people are using the AI music generation service Suno.

Not long ago, I also attended an AI composition seminar. The top LINE or track making created by AI, as well as the male-to-female and female-to-male vocal conversion features, are truly amazing. However, the part that impressed me the most was, surprisingly, the lyric delivery power.

Whether you input Hangul lyrics, English lyrics, or even K-pop style lyrics mixing Hangul and English, the closeness to the melody is quite good. Even when somewhat direct or awkward sentences are inserted, they sound natural. Even lyrics that lack so-called "resistance" sound quite convincing.

At first, I was amazed by the technological advancement. But after listening to the results many times, I had a different thought. AI singing sounds like a student who has been taking lessons for at least a year or more. It was not just a matter of pitch or rhythm. Above all, the pronunciation flowed naturally.

At that moment, the scene of vocal lessons came to mind.

There is a part that surprisingly takes a long time for students who are learning to sing for the first time. It is pronunciation. More precisely, it is the flow of pronunciation.

The pitch is correct and the basic framework of the rhythm is right, yet something feels awkward. The rhythm is correct, but it lacks sophistication. When listening to the song, the notes seem to play separately, and the words are cut off. It is as if a dish was cooked with good ingredients but the knife work was rough.

I call this the flow of pronunciation (flow).

Especially, Korean is composed of relatively distinct syllables as segmented sounds. The shape of the letters also has the initial consonant, vowel, and final consonant gathered like a single block, so there is a tendency to clearly distinguish syllables when speaking. This is even more so when the final consonant, or batchim, is attached quickly. In contrast, English words and sentences flow together through liaison, as if writing characters horizontally.

In singing, a technique is needed to appropriately connect the boundaries of syllables. Even when singing the same sentence, a skilled singer sounds like a single flow, while a beginner feels like the syllables are cut off one by one. That is why we unconsciously feel that the song is "sophisticated" or "skillful."

What is interesting is that AI implements that point very well.

Of course, AI has never loved at Navy Headquarters nor has it experienced a breakup. It does not know excitement or loss. It cannot feel emotions. Nevertheless, the emotion or message of the song is conveyed more naturally than expected.

This makes me think.

When conducting vocal lessons, we often emphasize emotion when talking about pronunciation. Put emotion into it. Put your sincerity into it. That is not wrong. However, feeling emotion and conveying emotion may be slightly different issues. Even if you sing with deep emotion, communication is not complete if it is not conveyed to the listener. Ultimately, emotion is conveyed through the context of words. And in that process, the connection of pronunciation plays a more important role than one might think.

Listening to AI songs, I realized anew. When we speak in daily life, we do not speak by cutting off each syllable like an ARS guide voice. We connect words and sentences and convey them as a single flow. Singing is similar.

People who are learning to sing for the first time focus on pitch and rhythm. Of course, that is important. But once they get it to a certain level, what is needed in the next stage is connection. The connection of notes. The connection of words. And the connection of meanings.

AI learns from the singing data of countless people. Looking at AI that has learned from this observer's perspective, I am reminded of the most basic fact again.

That what we hear in songs is not a simple list of syllables, but a fact of context and flow.

<© STARNEWS. All rights reserved. No reproduction or redistribution allowed.>

*This content was translated by AI.