Games · The Signal
Franklin AI turned more than 35 AI-generated 3D models into a kart-racing game
Franklin AI is demonstrating one of the most useful applications of AI-generated 3D assets: taking the models out of a preview window and putting them to work inside a game.
Reporting updated Sep 23, 2026
Franklin AI is demonstrating one of the most useful applications of AI-generated 3D assets: taking the models out of a preview window and putting them to work inside a game.
What Happened
It's easy to generate a model of a racing kart. Making that kart look right alongside dozens of other assets inside a game takes more work. Franklin AI decided to test that process. In a kart-racing prototype, the creator used Tripo Smart Mesh to replace basic placeholder visuals with a collection of textured 3D models. The experiment produced more than 35 assets, according to the creator, with Claude Code helping integrate the results into the game. Franklin also reported spending less than $14 on the generated asset set. That figure comes from the creator's demonstration and isn't an independently measured production cost. The accompanying video includes a Tripo discount promotion, which readers should understand when evaluating the cost claim.
Why It Matters
For independent developers, the real value of an AI-generated model emerges when it works inside the game. Scale, geometry, materials and collision all matter once an asset leaves the generation tool. Franklin's experiment makes that production handoff the story.
VEYR Angle
A good-looking 3D model is a starting point. A collection of models working together inside an actual game is something more useful. The interesting development is how much of that journey an individual creator can now attempt with AI assistance.