Helmet 3D model using Gemini and Tripo Ai

For the Red Bull car park drift event, I was responsible for designing the entire event out this focuses on the helmet-themed photobooth. The challenge was that the concept required using the athlete’s actual helmet design, which is highly customized and not readily available as a clean reference. Due to time constraints, recreating it manually wasn’t feasible.

To solve this, I built a workflow combining reference gathering (google), AI image generation (Gemini), and 3D generation with AI (Tripo AI)

First, I collected multiple images of the athlete wearing the helmet from different angles. Since no isolated product shots existed, I used Google Gemini to generate images of the helmet. I generated each view individually (front, sides, back, and top) because prompting for all views at once often results in inconsistent or merged outputs like an image collage I got when I first tried.

A key challenge was preserving accurate details, especially text on the sides. To address this, I manually verified the athlete’s name and fed that back into the prompts to refine the results.

Once I had a complete set of views, I moved to Tripo AI and used its multi-image input feature. This was important because the helmet design isn’t exactly the same thing on both sides, and single-image generation would have produced inaccurate model.

After processing, I exported the model as an OBJ file and imported it into SketchUp, where I integrated it into the overall photobooth design.

This workflow allowed me to quickly generate a highly accurate 3D asset from incomplete references, saving significant time.