Napkinmatic is a robust spatial computing platform connecting the real world to AI and back again in AR. The name derives from how the platform can turn your proverbial napkin sketch wireframe into a working website or app (GPT4 multi-modal), with several design sets to choose from (ControlNet), all augmented back on the napkin, so that you can interface with the app "in real time, in reality". The platform features an opensource framework which allows anyone to swap out the model with the latest SOTA by updating endpoints, as well as a built-in javascript engine for AppMode.
graph LR
I1[Prompt]-->B[**ControlNet**]
I3[Image]-->B-->C[vivid napkin design!]
graph LR
I1[Prompt]-->I2[**Prompt2Prompt**]-->B[**ControlNet**]
I3[Image]-->B-->C[vivid napkin design!]
graph LR
J[AppMode invoke]-->K[3D Sticky Notes Mode]
K-->L[3D model creation rectangle]
K-->M[Texture creation buttons]
M-->P1[Prompt]-->Q[Model Creation Method **FacedSculpt**]
M-->P3[Image]-->Q-->C
L-->I1[Prompt]-->B[**TEXTurePaper**]
L-->I3[3D Model]-->B-->C[original textured mesh!]
We raycast paintball splatters on "napkin" in our GDC23 AIGames Hackathon entry PaperSet Multiplayer and Instagram short here and it won 2 prizes, enough to cover SF rent for a month!