I don’t play any instruments, but now I can compose a full song from scratch
In the past, I co-authored lyrics for a couple of songs that were recorded in a professional studio, but I never thought I could compose a full song on my own—until now. Here’s how I make the song and its video. It’s a lot of fun!
I draw inspiration from what I read. First, I decide on the main idea the song will communicate, which shapes the title. Then I work on the verses and the chorus. The chorus repeats the main idea in a catchy way, while the verses enrich it from different angles. I iterate verse by verse, using tools that help me write stanzas with consistent syllable counts and rhyme patterns.
When the lyrics feel solid, I use a generative AI music tool to create draft songs from them. My prompt sets the musical style, vocals, rhythm, and tempo, and it describes each section of the song and how its speed and mood should change. If I did a decent job with the meter and rhyme scheme, the lyrics need little adjustment, usually just small tweaks so the words fit the melody. Sometimes I rewrite whole lines, and occasionally an entire verse.
Once the words fit the music, I experiment with different styles. It’s amazing how well-written lyrics can suit a variety of genres. I’ve produced one song in rock, merengue, metal-rap, and R&B, and every version sounds great.
After settling on a style, I regenerate the song dozens of times, making small variations to the speed, mood, intro, bridge, and outro. I quickly trim roughly a hundred versions to a handful of favorites, then narrow those to the top two or three. At that point, I live with them for a while, listening while driving, running, or walking, before I make a final decision. Occasionally I ask a few trusted friends for their opinion. Eventually I pick “the one” and produce a high-fidelity version.
From there, I use another set of tools to generate short videos of about 10 seconds each, typically one per line, or four per verse. Each video prompt aims to show what its line communicates within the context of the whole song. Sometimes the line itself works as the prompt, but often I have to adjust and enrich it to get the result I want. This part is super fun. It’s like playing Pictionary in reverse: I describe in words what I want the tool to generate, so the visuals creatively convey the intent of the lyrics.
As the clips accumulate, I stitch them together manually, using the audio track as the base and adding the lyrics as text overlays. Once I have a full first draft, I regenerate specific segments to improve their expressiveness, quality, color, brightness, or emotional and intellectual impact.
After many hours of hard, creative, and rewarding work spread over days or even weeks, I finally have the end product: a video that visually expresses what I set out to say through the lyrics, enriched by music I enjoy. I hope some of you enjoy it too!