November 28, 2022

The AI Art Revolution

If you’ve been online at all recently you’ve probably noticed the recent trend of images generated by artificial intelligence. These AI programs work by feeding a prompt to an algorithm which then renders what it thinks you’re describing. 

AI programs are trained on incredibly large datasets of reference images and use a process called Deep Learning to recognize what defines the individual elements of each image and the context in which each element is usually seen. It’s able to take these data points and apply them to the input you provide in a text prompt describing the content and style of the desired image.

For example, I created the image at the top of this post by giving an AI program the prompt: “An oil painting of an orange tabby cat wearing a spacesuit, dramatic lighting.” The AI recognizes the words “orange tabby cat” and “spacesuit” and has an understanding of what visual aspects are characteristic of an oil painting, based on its training with hundreds of millions of reference images. It generated three variations from my prompt.

From here, I decided to take the third variation and use a function called “Outpainting” to extend the image. Giving the AI more prompts in the same style, I added ground below the cat and planets in the sky. Theoretically this process could be repeated indefinitely, creating a larger and more complex image with each iteration. 

The potential applications of artificial intelligence are endless. As an artist, I see programs like this as a natural next step in augmenting our creative bandwidth. Now any picture within your imagination can be visualized within seconds with this incredible tool. It creates a great jumping-off point for artistic projects and often can stand alone as its own work. On top of all this, it’s still only in its infancy. Similar AI are being developed to apply this method to create videos and even 3D objects for use in virtual spaces like video games and The Metaverse. It will be an exciting next few years for this technology. 

If you want to be a part of this new AI art movement, here are a few resources you can use to create your own visuals. 

1. DALL-E 2 

Operated by Open AI, DALL-E 2 is the second iteration of the original DALL-E AI released in January 2021. It allows you to not only generate images from prompts but also use outpaintng to create more around the image as well as inpainting to make edits inside the image. 

2. Nightcafe 

Nightcafe is an Australian based studio on a mission to democratize art creation. In 2019 its founder Andrew Russell set out to create more personalized art to hang on his walls and ended up creating one of the most popular AI art generators out now. 

3. Artbreeder 

Artbreeder takes a number of different approaches to generating AI art. Using a feature called Collager, the user is able to draw out shapes on the screen which will define the general composition of the image. Then they can enter their prompt and the AI will project this prompt onto the shapes, creating the image. Another feature called Splicer allows the user to mix images together to make a hybrid image.

4. Lensa AI 

If you’ve seen the recent trend of people posting artistic avatars of themselves online, its coming from Lensa AI. Made by Prisma Labs in Sunnyvale California, this app has been around since 2018 but has seen a massive increase in downloads with their new “Magic Avatar'' feature. Lensa AI allows you to upload 10-20 reference images of yourself and the AI will generate a batch of “Magic Avatars” in various styles, which you can purchase for 8 to 15 dollars depending on the size of the batch. 

5. Wonder AI Art Generator 

Wonder AI is an app which allows users to create images from prompts from their phone. The app has a bunch of different styles that you can select for the aesthetic of your image or you can simply select no style and describe the style you want in the prompt. I’ve been using Wonder AI for a few weeks now and while it's not quite as advanced as some of the other AIs on this list, it's still been a great resource for generating AI art on the fly.