The project is a generative NFT that uses a decentrally hosted AI trained on trippy images with a web app interface for minting. Users can mint the NFT through the dApp and their transaction and token information will be used to seed parameters to an AI that is tuned for producing psychedelic art.

SocksOff showcase

How it's made

The project uses @Chainlink VRFs in the contract to make the random numbers that will inform parameters of the generative model at time of minting. Brownie's nft-mix was used as a starting point for the project as well. Currently we just use a script that makes fractals with unique parameters depending on the number, but we want to migrate to the generative adversarial network (GAN). Image data for the fractals is hosted with @ipfs. The GAN is going to be trained on a custom image dataset we made from webscraping mandalas and generally trippy images. The idea we had for decentralizing it was to encrypt the model using @Nucypher, and to have the model's script stored on-chain and encrypted. Then a content delivery network would communicate with our front-end to run the stored on-chain model and generate images based off of the parameters an end user received for their token. The webapp uses @Moralis as a middleware.