Dall-E 2 by Open AI, Google Imagen, Midjourney, and even new Open Source competitors are AI tools, that can generate images from textual descriptions. Just from a description of objects, scenery, and art style, you can generate thousands of unique pictures.
AI-generated images work by taking text input and creating an image that resembles the description. The tool uses a neural network to interpret the text and generate an image that matches the description. The neural network is trained on a dataset of images and descriptions. The more data it has, the better it can interpret the text and generate an image that matches.
Above, you can see five results of the prompt "a white table in a black room, 3d rendering". Next to describing objects and art style, you can specify aspect ratios and how much "freedom" the AI should have while creating the images. You can define if it should stick to the description as well as possible or get more abstract. Below you can see an example of that.
This means that instead of searching for stock photos that approximate what you're looking for, you'll be able to simply describe what you need and have an AI tool create it for you.
The implications of this are huge. For one, it will make it much easier for businesses and individuals to find the exact image they need, as they won't be limited by what already exists.
It will also make it possible to create completely custom images, which could be used for everything from marketing campaigns to product designs. And since these tools will be constantly learning and improving over time, the possibilities are endless.
This technology is still in its infancy, but it's rapidly evolving. In a few years, it's likely that AI-generated images will be the norm, not the exception. So, if you're in the business of creating visuals, it's time to start learning about image AI.
Fun Fact: This article was written with the help of GPT-3 by OpenAI. But we’ll get to that in one of our next blog posts.