Art generation principles
Last updated
Last updated
First, let's understand the basics: the AI receives a picture and some text as an input. The output is a visualization of our text based on the resulting image. This is how it looks in practice:
In this example we use Starryai to generate art, but we will use different services and models.
The art generation model of #WEAREONE is based on AI, news, trends and auction winner phrases. It is best to start by visualizing the whole process of image generation (we also call it ArtChain, since each image contains data from all previous images):
The step-by-step process of building ArtChain can be described as follows:
For the first iteration, we take our secret Zero Genesis image, words from founders with trends and use all of them to generate Genesis picture №0;
Then, we take the Genesis picture of the current day and one piece of news and use it to generate AI picture №1;
AI picture №1 and another piece of news are used as the basis to generate AI picture №2. We repeat this step five more times until we get the final picture of the day.
The picture of the day is converted into an NFT (the previous pictures are not NFT's) and goes to 20h auction.
The auction winner uses his phrase to describe the day.
The auctioned picture, the winner's phrase and the trends of the day are then used to form a Genesis picture for the next day.
The next day starts with a Genesis picture and one piece of daily news.
👀 A secret cultural ingredient is added to the Artchain at a random iteration.
Important notes:
At the initial stage only the final picture of the day will exist in the form of NFT. Perhaps in future stages we will put the entire image generation process on a blockchain or other decentralized and fault-tolerant data structure.
We will take macro-region news from aggregators manually, but we will pay attention to the relevance of the data (the number of social interactions, the number of views and other objective metrics).
At the initial stage, we will use various models of the Starryai and Midjourney for image generation, but in the future we plan to test other models.
At the current stage of technology development, we cannot fully automate the entire process, so human factor is going to be involved in the project. Multiple members of the DAO will participate in the image generation process. Image generation will not depend from just "one admin", which will allow ArtChain to continue working daily. Our goal is to fully automate the entire project so that news selection and image generation processes are decentralized, and all the intermediate stages of generating the final NFT are also available in a publicly accessible and fault-tolerant database (not necessarily a blockchain, but some kind of a decentralized network).