fabFab

Inspiration

One participant had struggle finding game assets, known as fabs for a 3d environment and we thought we can create these fab assets with gemini ai.

What it does

Firstly creating a system where we can:

Input a prompt along with an image (optional) --> generates a ".obj", --> return to the user with modified features --> then the user can adjust --> will remodify the ".obj" to the users demand and return.

In the website, we used three.js to show a sample of the .obj file which can be downloaded afterwards.

How we built it

Backend: FastAPI, google gemini, and google adk.

Frontend: Typescript (lovable), Three.js, and axios to connect to backend (incomplete).

Challenges we ran into

Communicating with gemini to give accurate results, but limited to tokens. Struggled with fitting a gemini's agent to fit our architecture. Postman commands would stall and fail calling gemini.

Accomplishments that we're proud of

Developed a secure backend in case of failure. Developed a friendly frontend that supports 3D obj using Three.js.

What we learned

Although AI is amazing, AI has trouble developing 3D assets at higher levels unless having a lot of tokens. Similiarly to how AI struggled with artwork and now excels at it, maybe we will explore it in the future.

What's next for fabFab

Completing it lol. Definitely optimizing our token usage so we can have more refined 3D models. Likely move from obj to other files like fbx. Lastly include textures for the 3D objects for complete fab design.