The prompt for this challenge was “Bubble Wrap.” I’ve always enjoyed bending prompts just enough to surprise people, so instead of taking it literally, I leaned into a phonetic twist—Bubble Rap. It felt like a playful loophole that still honored the spirit of the challenge.

As it turns out, writing something that can actually be rapped is harder than it sounds.

My early drafts focused on simple end rhymes, building the story piece by piece until the bubbles and final twist emerged. On paper, it looked fine—but when read aloud, it felt clunky. I shifted my attention to syllable counts and cadence, trying to make the rhythm more natural. That helped, but something was still off.

Eventually, I stopped fighting it.

Instead of forcing it into a strict rap structure, I leaned into something closer to a playful, almost childlike rhythm. That shift unlocked the tone. The exaggerated drawl in the line “Thrrrrreeeeee in aaaaalllllll!” became a turning point—less about technical precision, more about performance and personality. It’s the kind of piece that’s more fun to say than to simply read.

The ending carries a small personal reference. Years ago, a family friend—just a kid at the time—casually said, “That poor car, it’s red.” The delivery and absurdity of the statement sent my brother and me into hysterics. It’s one of those oddly specific memories that stuck with me, resurfacing at random moments. That tone found its way into the final line:
“That poor bubble, he’s dead!”

When I submitted the piece, I included a brief note admitting that I’m not much of a rapper—but I gave it a shot anyway.

Experimenting with AI Performance

Since the piece was written with a rhythmic, performative style in mind, I decided to experiment with AI-generated audio and video to bring it to life.

Getting the right sound proved challenging. I aimed for an ’80s-style rap delivery, but most outputs missed the mark. The closest attempt was still imperfect—especially with pronunciation. The word “jokester” consistently came out as “joke-es-ter,” which occasionally drifted into something that sounded unintentionally off.

The video experiments were even more unpredictable.

I uploaded a photo of myself, along with my 100 Word Stories logo, and described my physical appearance to help guide the model. While this approach has worked reasonably well for generating still images in the past, video introduced new complications. The main character’s face shifted from scene to scene, rarely resembling me for long. Lip movements didn’t align with the lyrics, and at times it felt like watching a muted television while unrelated audio played in the background.

There were also plenty of unexpected artifacts. Early versions included multiple Christmas elves appearing and disappearing throughout the scene as the jokester. Later versions removed them almost entirely. The “jokester,” as I imagined him as a separate character, never really materialized visually, despite being central to the story. The logo I provided was simplified into a gold medallion, losing much of its original detail. The video was cut off at 60 seconds.

Subsequent attempts improved certain aspects but introduced new quirks. At one point, the character appeared to be holding a medallion that was also hanging around his neck. In another, he performed with a microphone in each hand. Typewriters appeared frequently as I asked it to lean into my 100-word story logo that featured a typewriter. The second video cut off at 76 seconds.

On my third attempt, I monitored the rendering process more closely. The system built the video incrementally, generating short clips that were loosely connected. Occasionally, a frame would briefly resemble me before shifting again. The final version ran longer—79 seconds—and included a more complete ending, though it still cut off just slightly at the very end.

The entire process cost roughly 8,000 credits for one video—an interesting experiment, though not one I’d call fully successful.

Reflections

What stood out most wasn’t the imperfections—it was how close these tools are getting, and how much they still depend on human intent, timing, and interpretation. For now, we adapt to the tool’s quirks, rather than the other way around.

In the end, the written version remains the strongest form of the piece.

It lives somewhere between a rap, a rhyme, and a short performance—exactly where it feels most natural.