Creating an AI influencer is a fascinating blend of creativity, technology, and strategy. As I continue developing BoxxBoxx, my blonde, race-suit-wearing Formula 1 AI influencer, I want to share my progress, challenges, and learnings.
Why Formula 1?
The decision to focus on Formula 1 was straightforward for me. Unlike other sports that don’t particularly interest me, I never miss an F1 race. This passion means I understand the sport well, can create relevant content more easily, and genuinely enjoy the process.
When designing BoxxBoxx, I considered various character options. Should she be realistic? Cartoonish? Perhaps even an anthropomorphic animal? Ultimately, I decided on a realistic-looking woman while ensuring transparency about her AI nature.
Designing BoxxBoxx
The name “BoxxBoxx” (with double x’s) came from ChatGPT’s suggestion, based on the term “box box” that race engineers use over the radio to call drivers into the pits. I wanted something catchy and F1-related without infringing on any trademarks.
For her appearance, I started with ChatGPT, which I’ve found surprisingly good at maintaining visual consistency. I designed her as a blonde woman wearing a black racing overall and cap, featuring a custom “BoxxBoxx” logo.
While initially pleased with ChatGPT’s results, I later discovered Google’s Nano Banana image model through Gemini. This tool excels at modifying existing images without changing the original subject, allowing me to create various perspectives and expressions of BoxxBoxx.
Nano Banana has significant limitations, though. The output resolution is extremely low, making text on clothing barely readable in wider shots. To address this, I’ve used Topaz for upscaling key images to higher resolutions.
Building BoxxBoxx’s Online Presence
I’ve established BoxxBoxx across multiple platforms:
- A YouTube artist channel
- A Spotify artist page
- A TikTok artist page
- A standard Instagram profile
The content is minimal so far, especially on Instagram where I only have about five followers. The accounts need optimization, better descriptions, and more engaging content.
Creating F1 Race Recap Music
My unique content angle is creating hip-hop/rap songs that recap Formula 1 races. This format allows for straightforward race reporting, and ChatGPT excels at writing rhyming lyrics for this purpose.
The production process works like this:
- I grab the race report from a F1 news website
- Feed it into my custom GPT that converts it to song lyrics
- Upload those lyrics to Suno (an AI music generation tool)
- Apply a custom style I’ve developed for BoxxBoxx
- Add a filter in Adobe Audition to give the track more fullness
- Create AI-generated F1 visuals using Kling and VEO3
- Combine everything in Adobe Premiere and distribute to all platforms
I tested this workflow with the recent Azerbaijan Grand Prix in Baku. After watching the race on Sunday, I retreated to my office and produced the entire song and video within an hour.
First Challenges and Learnings
Despite having everything technically in place, growth has been minimal. On YouTube and Spotify, with zero followers, my content isn’t gaining traction. Even on TikTok, where follower count matters less due to their algorithm, my videos are only reaching a few hundred people.
I’ve identified several issues to address:
The hip-hop genre might not be ideal for F1 fans. While tracks like “Super Max” (about Max Verstappen) have millions of streams, they’re upbeat party songs played at F1 events and viewing parties. My rap style doesn’t naturally fit those environments.
The TikTok videos aren’t grabbing attention in those crucial first seconds. The AI-generated F1 footage I’m using is generic rather than specific to the race being recapped, creating a disconnect. The songs themselves aren’t structured to hook listeners immediately.
Future Plans and Refinements
I’m considering several adjustments to my approach:
Music genre evolution – I’m exploring “Hip House,” a fusion of rap and dance music that might better suit the F1 fan environment while still allowing for race recap lyrics.
Content expansion – I’m working on an album about Lewis Hamilton’s career, believing his fans might be more receptive to hip-hop. I also have a tribute to Senna ready to publish, potentially timing its release with McLaren winning the constructor’s championship.
Process improvements – Creating content immediately after races competes with family time, so I need a more sustainable workflow.
Expert conversations – I’m scheduling podcast interviews with music and F1 experts, including Patrick Kicken (who works with Suno) and potentially Alexander Stevens from Grand Prix Radio.
Looking Forward
Despite the slow start, I remain optimistic. I’ve had success with another AI music influencer that’s gained hundreds of thousands of streams across platforms. I’ve learned valuable lessons about Spotify’s algorithm, TikTok growth strategies, and AI content creation tools.
In future podcast episodes, I’ll dive deeper into specific aspects of this journey – from the technical details of tools like Elevenlabs (for creating BoxxBoxx’s voice), to content formats beyond music, including vlogging. While I haven’t gained significant reach yet, the foundation is in place, and I’m excited to build upon it.
The BoxxBoxx experiment is just beginning, and I can’t wait to share more of this journey with you.
Listen to this article
This article is based on one of my Dutch podcast monologues. You can listen to it in the episode below. It is part of the AI Influencer Lab series, in which I take my listeners through the steps of developing an AI influencer for a Formula 1 audience.
