This is a demonstration how to combine multiple classical graphics approaches with AI generated art to create a 3D web-based experience. The approaches used include text to image, image to video, text to 3D, and video to 3D using a technique called gaussian splatting.
To run this locally and experiment with modifying the code, clone this repository, and run the following commands from the root of the folder.
npm i
npm run dev
\
Here are some links to the tools used for creating this:
Dalle3 (running within ChatGPT). Text to Image model (Used to generated stained glass and other textures)
Blockade Labs A tool for creating 3D environment maps using a combination of prompting and sketching.
Magnific.AI An image upscaler and enhancer
RunwayML A text to video and image to video tool
Krea.AI A tool that allows the user to compose various inputs such as webcam capture, screen capture, and shapes, with a prompt and outputs images based on these in realtime
LumaWebGL Library A library that renders scenes captured from the luma app using webgl.\
[Genie] A text to 3D model tool. (Used here for the candles and sculpture of the woman)
Blender A free and open-source 3D content creation tool.\
chapel model by adamleescott