Back in 2018, renowned auction house Christie’s became the first establishment of its kind to sell a piece of art generated by artificial intelligence. Entitled “Edmond de Belamy, from La Famille de Belamy,” the artwork hung in New York City among an Andy Warhol print and a bronze work by Roy Lichtenstein. It was sold for a whopping $432,500 — reportedly more than double the price of both its neighboring pieces combined, and some 40 times more than what Christie’s initially estimated it to be worth.
AI Art Generators to Know
- DALL-E 2
- Stable Diffusion
Since then, AI-generated images have become one of the hottest areas of innovation in the artificial intelligence space. People have used this technology to create everything from illustrated children’s stories to award-winning artwork; and the recent development of dozens of AI art generators has made it possible for basically anyone to create unique, high-quality images with nothing more than a quick text prompt or an uploaded photograph.
Thanks to the latest advancements in AI art generation, people can create images virtually out of thin air. People no longer need natural artistic ability or technical talent to create incredible art.
What Is An AI Art Generator?
AI art generators are software that use artificial intelligence and its many facets (machine learning, neural networks, and more) to automatically create images. They often come in two forms — text to image or image to image — and both are fairly self explanatory.
What Is An AI Art Generator?
AI art generators are software that use artificial intelligence to automatically create images. These tools often come in two forms: text to image or image to image.
Text-to-image generators are able to create unique images based on a user’s text input, which is also commonly referred to as a prompt. These AI tools are trained on massive data sets that pair text and images. Meanwhile, image-to-image generators allow users to upload an existing photograph or image, and then edit or transform it using the power of artificial intelligence.
How Do AI Art Generators Work?
When it comes to text to image, however, most tools start with researchers feeding them millions and millions of images. They also annotate these data sets, providing the machine with a text reference. Then the device is tweaked, calibrated and trained until it can consistently recognize those photos. After a while, it can begin to parse the difference between varying images, and generate its own with some accuracy.
From there, most text-to-image generators function in a similar way on the user’s end: A person enters a text input into the tool (“bunny in a raincoat painted in the style of Frida Kahlo,” for example) and the tool spits out a creation that closely represents the image described. This new image has been generated by an AI algorithm, and results can vary widely if even a small change is made to the prompt.
Image-to-image generators are a bit more straightforward, and often use a process called neural style transfer — an optimization technique that takes a content image (uploaded by the user) and a style reference image (artwork by a famous painter, for example) and blends them together so that the resulting image looks like both the content image and the style reference image. In the end, a user can turn a photograph of their dog into a Renaissance-style painting, a selfie into a pencil sketch, and so on.
In general, because there is so much going on under the hood with generative AI that even some experts don’t understand, there is a lot of uncertainty as to what kinds of images a tool will spit out at all. At the same time though, many AI art generators use third-party technology behind the scenes, meaning there could be dozens of AI art companies using the exact same AI system under the hood. And the tech stack that powers AI art generation is often accessible to the general public, allowing for even more innovation.
While impressive, this technology has not come without its fair share of pushback (particularly among artists), and it has forced society to reexamine what art can be, and whether or not creativity is inherently human.
Still, interest in AI generated art has never been higher, and these tools are pushing the boundaries of what’s possible.
20 AI Art Generators to Know
A sort of cross between an art editor and a text-to-image generator, Artbreeder has many capabilities. Users can draw rudimentary images and shapes, type in a description of the kind of image they want to create, and Artbreeder can bring it to life with its artificial intelligence and machine learning capabilities. Users can also edit and enhance existing images, generate multiple iterations of a given image, and even turn still images into animated figures. The platform also hosts a fairly active community of creators, who can show off their AI-generated art and collaborate with each other.
Unlike many other AI art generators, Big Sleep is Python-based, so it is geared toward more creators with coding experience. Although the images are not quite as advanced as some of the newer AI technology available today, creators can use Python script on Big Sleep to generate realistic AI-generated images from scratch. Big Sleep is open source, free to use, and available on GitHub.
Formerly called DALL-E mini, Craiyon is an AI model that can generate images from any text prompt, and it is totally free to use. The model, which is constantly improving, is trained by looking at millions of images on the internet and their associated captions. Over time, Craiyon learns how to generate images from a simple text prompt. Some of these concepts are learned from memory, while others are entirely unique and have never existed before, and are made possible by combining multiple concepts together. Because the model was trained on unfiltered data from the internet, it may generate images that contain harmful stereotypes or biases, but the extent and nature of these biases are continuing to be looked at and corrected by Craiyon’s development team.
Created by nonprofit research company OpenAI, DALL-E 2 has become one of the most prominent names in AI art generation over the last year. Its ability to create accurate and realistic images with nothing more than a short text prompt is made possible thanks to a process called diffusion, which begins with a pattern of random dots and gradually alters that pattern into an image once it recognizes specific aspects of that image, the company explained on its website. In the end, users with absolutely no natural artistic ability can create everything from a painting of a fox in the style of Claude Monet to a photo-realistic image of a polar bear playing guitar. Like its predecessor DALL-E, DALL-E 2 began as a research project, and is now available to all with both free and subscription options.
Developed in 2016, DeepAI is among the oldest AI art generators out there now. It is fairly simple to use: All you have to do is type in an idea for an image and select the style you would like the image to be in from the platform’s library. The platform is then able to create the image in a matter of seconds thanks to the power of artificial intelligence. DeepAI can also be used to colorize existing images, and detect specific objects within images through an offering called Zendo. This can be especially useful for self-driving vehicles and medical imagery.
Deep Dream Generator
After training itself with millions of images using advanced AI algorithms and techniques, Deep Dream Generator uses neural networks to generate new images based on existing images — making it different from the various text-to-image AI art generators out there. Users simply upload an existing image, choose whatever art style they would like the output image to be in, and then a new image is produced. The platform does have a text-to-image offering as well. While some of its creations look photo-realistic, Deep Dream Generator is perhaps best-known for the more psychedelic, surreal images it produces.
Dream by Wombo
Dream by Wombo allows users to transform existing photos into cartoons or faux paintings, as well as turn words and phrases into unique works of art via a complex algorithm. All of these creations can then be turned into NFTs. This tool was created by Wombo, which allows users to create lip-syncing deepfakes of themselves, friends and even famous people.
Fotor is an AI image generator that can turn photographs into any art style, from sketches to oil paintings to water colors. The platform has also carved out a space for itself in the creation of NFT artwork, allowing users to quickly create AI-generated digital art, export it as an NFT, or non-fungible token, and then sell it on NFT marketplaces.
Jasper allows users to get as detailed as they want with their creations, offering not only a text box for prompts, but also various drop-down menus dictating style, medium and even mood. In addition to its AI art generator, the company can also create custom written copy, providing users with AI-generated blogs, social media content and more with nothing more than a short prompt. Jasper is just one of many players in the automated content creation game, including Writer, GPT-3 and ChatGPT. The company recently rolled out a chatbot function as well.
If you’ve been scrolling through your Instagram or Twitter feeds over the last several months, odds are you’ve come across at least a handful of artsy profile images shared by people in your network — courtesy of Lensa. The AI art generator’s recent Magic Avatar feature has taken social media by storm, using an advanced AI model to generate artistic versions of selfies based on visual blueprints users provide. The platform was created by Prisma Labs, with the goal of bringing mobile photography and video creation “to the next level” using neural network, deep learning and computer vision techniques, according to the company’s website. With Lensa, users can make minor edits like background blurring and filter applications, as well as create entirely unique renderings based on a single photo.
At first glance, Midjourney is similar to a lot of other AI art generators on this list: It is a text-to-image tool that spits out creations from scratch in a matter of minutes. But the art it can produce is so realistic and impressive that it is the only platform of its kind to win an actual art competition, sparking both awe and controversy. Midjourney is now available in beta, but access to it is restricted to a select few. To use it, you have to apply to its official waiting list. Once accepted, you will receive an invitation to its Discord server, through which you can interact with the tool.
NightCafe is one of the most popular AI art generators on the market today, and is said to have one of the biggest repositories of algorithms and features. Similar to other tools, NightCafe creates original pieces of art from users’ written prompts. One unique feature, however, is that it runs on a credit system. The more credits a user has, the more images they can generate. And credits can either be purchased or earned by taking part in NightCafe’s community.
Like many other platforms, there are two ways you can generate art on Photosonic. One is by entering a text prompt that describes the type of image you want to create, the other is by converting an existing image into an AI art image. The model Photosonic uses is based on latent diffusion, meaning that it gradually transforms a random noise image into a coherent image that matches a given text, according to the company website. Users can control the quality, diversity and style of the generated images by adjusting the description and re-running the model. While users have full rights to the images they create on the platform, so too does the company and its other users. This means all Photosonic users can freely reproduce, display, distribute and create derivative works based on any images on the site, without any royalty or attribution requirements.
Point-E is the latest release out of OpenAI, which promises to be like DALL-E 2, but with 3D modeling. Three-dimensional modeling is vital across a variety of industries and applications — from CGI effects in movies and video games, to NASA’s moon crater mapping missions to the creation of the metaverse. But creating photorealistic 3D images remains a costly and time-consuming process. That’s where Point-E comes in. The platform can produce 3D point clouds directly from text prompts by leveraging “a large corpus” of images paired with text, while the model is trained on a data set of images paired with 3D renderings, the OpenAI team wrote in a paper published last month. “To produce a 3D object from a text prompt, we first sample an image using the text-to-image model, and then sample a 3D object conditioned on the sampled image. Both of these steps can be performed in a number of seconds, and do not require expensive optimization procedures.”
Runway uses machine learning models to generate unique images based on text prompts. But its main focus is on using artificial intelligence to create animations and edit videos. For example, the tool can be used to remove and replace a background from any clip without the need for green screen technology. It also allows users to create their own animations and 3D models and use relative motion analysis to improve motion graphics.
Developed by digital assets marketplace Creative Fabrica, Spark has a suite of AI tools, including a text-to-image generator, a print creator and a copywriter that can be used to automatically create texts for articles, books, advertisements and more. Like other digital assets on the platform, users can post their creations for paid use by other members, reportedly making it the first generative AI platform that also allows creators to make money.
Stablecog is an open-source AI image generator that allows users to create realistic images from scratch. The site also features a fairly active gallery with users’ creations, from a Pixar-style rabbit astronaut to a wide-eyed cat painted in the style of Vincent Van Gogh.
Stable Diffusion is an image-generating AI tool created by Stability AI. It has both text-to-image and image-to-image capabilities and is completely free to use. Plus, its license allows for the commercial and non-commercial use of the tool at no cost. And its underlying AI art model is used by some of the tools also on this list, including Deep Dream Generator and NightCafe.
StarryAI is an art generator that doesn’t require any input from a user at all. Rather, it can process images with a machine learning algorithm. One of the platform’s key selling points is that it acts as a free NFT generator, giving users full ownership over the images they create. Unlike many of the other tools on this list, StarryAI is available only as a mobile app for iOS and Android users.
With Visionist, users can apply AI art styles to any photo, including abstract filters, cutout portraits and more. The platform was created by faculty and graduate students at Parsons School of Design at The New School, and is used as not only a library and tool for creatives, but also a place to facilitate research into the relationship between creativity and AI. Visionist also experiments with AI-generated handwriting, AI-generated sounds, the transformation of words into unique patterns and more.