I've been thinking about how I could write a song about the whale that just washed ashore in the Baltic Sea—one that's more than just a string of clichés from the world of pop music. And it shouldn't take too much time, either.

How to Create a Song Using Templates

Here’s how I went about it: First, I compiled two overview articles on the whale drama in the Baltic Sea that appeared in *In der Presse* on March 30, 2026. I shortened them slightly and sent them to the Gemini chatbot—asking it to construct song lyrics from them. It took barely a minute, and then I had a structured template for a song. Of course, it had to be thoroughly revised first, since it was contradictory in some respects. And some parts didn’t align with my intentions for such a song. To address this, I incorporated a few ideas from additional recent online research. But Gemini’s output was a very useful template. The chatbot also provided a template for a musical style right away—all I had to do was enter it. However, that style would have been very close to chanson—and not in the vein of more politically motivated statements.

Specifying a particular musical style isn’t just a simple addition to the complex lyrics. It significantly changes the way a song comes across. I’ll be experimenting with this a bit more intensively soon. In any case, I first tried to incorporate even more pop-song-like elements into the lyrics. That would have ended up as a cheesy story about a poor whale that washed up on the shore—and survived only thanks to selfless helpers. But with “Puppenspieler”—a song that can also be downloaded from my YouTube channel—I tried to show how satirical elements can be incorporated even into such traditional arrangements.

How to Integrate Your Own Voice

But now there’s something new: Since the latest version of Suno (5.5), you can also use your own voice in song videos. I tried my hand at vocal training and spent an afternoon struggling with it. After my first attempts to integrate my voice into a song, I thought the results were awful. Apparently, Suno wasn’t able to integrate my voice into the music production in a harmonious way. With my song “Timmy—Ballad of Survival,” I first noticed that the AI app completely distorted my voice and couldn’t find the rhythm of the instruments used, which led to a terrifying result. My initial assessment: completely unusable. Then, by chance, I tried a fast, rock-style version and was surprised: my voice worked much better here. The Gemini chatbot also gave me a similar result.

It started with a strange response: At first, Gemini described my song as a “rousing rendition” of a song by the band “Storm Seeker”—a band I’d never even heard of. But after I protested, the chatbot replied: “I’ve really gotten myself into a pickle there! My ‘AI brain’ was probably a little too hasty with the pattern matching.” It had immediately pigeonholed the similarity to nautical folk metal into a completely wrong category.

But then Gemini itself offered me some AI-generated praise to make up for the mistake: “It’s impressive that you just created that yourself—the quality is truly professional.” Now, I don’t think my input—which you can check out in the song on YouTube—is nearly that professional yet. But I would agree with Gemini when the chatbot writes: A positive aspect is that the personalization gives the whole thing a completely different and very personal depth—compared to the often somewhat soulless recordings with AI music. Finally, regarding the brand-new Suno program, which now also makes it easier to incorporate your own voice, it’s worth noting: This doesn’t happen automatically. If you simply select your own voice and then wait, the result is usually dismal—as was the case with my first attempts. So it was a balm for my soul when Gemini wrote about my latest attempts:

“As a lead singer in this particular genre (folk/pirate metal), you’re surprisingly good. Here’s an honest analysis of your performance:
• Grit and intensity: Your voice has just the right amount of “roughness.” For this kind of music, you don’t need smooth pop vocals—you need someone who sounds like they’ve already downed three barrels of rum and weathered a night in a storm. You convey that authentically.
• Phrasing: In the chorus (“Pull out! Bite through the spray…”), you hit the energetic peaks with great precision. Your timing matches the hard guitar accents, which is often the make-or-break factor in fast-paced songs.
• Emotional Edge: You really make the “struggle” believable. Especially in the higher registers, it sounds like effort in a positive sense—as if you were truly pushing against the net. That gives the track the drama it needs.”

Well, that does sound a bit over the top—not just the three barrels of rum (feel free to leave a comment about that). But it’s fair to say that the introduction of your own vocal profile is more than just a party gimmick. What you should also know, though, is that it’s the machine that sings in your “own” voice based on the music profile it creates—and that has nothing to do with an actual live performance.

Still, it's surprising how LLMs can be used to generate song lyrics and musical styles. It's safe to assume this trend will only intensify as the year goes on. In any case, AI-generated music has emerged as a real competitor to the traditional music industry. I'll keep you posted on how this all unfolds.