Inspiration
We started by wanting to recreate the analysis environment from Detroit Become Human but realized a use case with a large gap in the market: Interior Design. With tools today either being 2d AI image generators or 3d models created by hand, there was nothing bridging the gap, so we built Stage.
What it does
You prompt your idea for a room and the chat asks you some clarifying questions to lock in the aesthetic. Then our AI generates a unique design for you. Then look for similar products to buy for your own home.
How we built it
We used gib files from free sites such as PolyHaven and SketchFab then rendered them using Three.js. We used Gemini to take a user prompt and generate a 3D layout based on the catalog and their metadata. Then we connected Gemini to Tavily to search for similar products on the web.
Challenges we ran into
The biggest issue was sourcing the furniture. Originally we planned to scrape the IKEA catalog but could not get the 3D models. So instead we chose to download open source 3D models, the tradeoff being not being able to buy the exact item. We solved this by product matching using Gemini Vision and Tavily to search the web.
What's next for Stage
The next steps for Stage would be to implement sharing features where users can collaborate, fork, and save their designs. Furthermore, we want to attach affiliate links to the product buying stage.
Log in or sign up for Devpost to join the conversation.