AI art is among the most striking ways to employ generative AI, which has caught the public imagination in dramatic fashion. But what is it, how do you use it and is it going to be a net positive for society?
Up to 60% of jobs in “advanced” economies will be affected by AI according to the IMF. It posits around half of these jobs will use AI to increase productivity, while the other half will be replaced by AI to some extent.
Will the same happen to artists? The original dream was AI would handle boring jobs, and let humans spend more time with creatively fulfilling endeavours. That hasn’t really happened, and we’re not sure how this one is going to end yet.
Generative AI may impact the livelihoods of musicians, artists and writers more than most. AI paintings and AI art can be generated in moments using text-to-image tools, and just about anyone can use this tech.
They are already incredibly powerful tools that let people undertake projects that might have been unfeasible just a year or two ago.
Here’s how it works, a look at some of the best AI art tools out there and what it may mean for the future of art.
What Is AI Art?
AI art typically refers to images made using generative AI tools. These create images of all sorts, of almost any style, on almost any subject, based on a text prompt.
In early generations of this technology, from 2022 and 2023, AI image generators would typically struggle with certain elements. Images of people would often have far too many fingers and teeth, for example, but the quality of output in 2024 iterations of the most popular generative AI models can be remarkable.
But is it really art? This is too big a question to answer in a brief explainer, but it is clear an “AI artist” typing in a few lines of prompt text should not be compared to a painter or illustrator who has spent years honing their craft. The skill involved is not in remotely the same ballpark.
One recent example of AI art being used in a high-profile way on the cover of Sarah J Maas’s best-selling novel House of Earth and Blood. This was brought to light by The Verge, but publisher Bloomsbury said its team was not even aware the image was AI generated when it was licensed from an Adobe image library.
It looks great on the cover, but did cause a heap of controversy among artists and fans back in 2023.
How Does AI Art Work?
You can ask an AI art tool to create an image of almost anything. Their intelligence is not just in creating images, but in determining what you intended in a text prompt.
You decribe the scene you want in words, and it can include terms like “in the style of,” with reference to specific artists, styles or periods of art. Photorealistic or cubist are just a couple of examples you might want to try to give AI generated art a go.
Some AI art tools can also use a source image as well as a text prompt, letting you turn one image into something completely different, while maintaining something of that original.
AI art is generated using a model that has been taught using a huge dataset. For example, OpenAI image generator Dall-E 2’s dataset features 650 million images, with linked text terms, pulled from the internet.
What Are AI Art-Generation Software Tools?
In the broadest sense, all generative AI tools could be considered AI art makers. The written word, music, static images and video can all be works of art, but “AI art” in common parlance today refers almost exclusively to AI-generated images. AI music generators are also less well developed at present.
These areas are starting to merge together, though. For example, the famous chat bot ChatGPT can be used as an image and text generator, because OpenAI’s Dall-E AI art tool has been integrated with the bot. However, this does require a paid OpenAI subscription.
There’s no real difference between AI art and an AI image beyond the text prompt used to make that image. It does not matter to the AI whether you want a picture of your local church in the style of a Picasso or a photo of Donald Trump riding a moped.
Images like that last example also make the ability to identify when images are AI generated and when they are not crucial. OpenAI has announced plans to add digital watermarks to Dall-E 3, both in visible and invisible form, but concedes this information can still be removed. There are also standalone AI detector tools, but most focus on text rather than imagery.
Top AI Art Generator Tools
The sheer heat of AI has seen a huge number of AI art tools pop up. Some examples include Stable Diffusion, Midjourney, Dall-E, Adobe Photoshop’s Generative Fill and Leonardo.
These are the best-known AI image makers. However, they require a paid subscription. Free-to-use tools at the time of writing include Deep Dream Generator and Dezgo. Both offer a good level of control over the end result even if they don’t use the very most advanced AI back-end.
Midjourney V6
Midjourney AI is one of the original AI art platforms, founded by David Holz. Its first iteration came out in 2022, while the latest V6 arrived in December 2023.
Basic access to this AI tool used to be free, but it’s now behind a paywall. Subscriptions start at $10 a month, rising to up to $120 a month for power users.
Its comprehension of text prompts is excellent, and the quality of its results are generally very good. However, Midjourney is far from the easiest AI art tool to use. You access it through popular chatting app Discord, which may seem odd given how well-known Midjourney is.
You will need to learn certain parameters to get the most out of it. For example, “— AR 16:9” means you want the output to have a standard widescreen aspect ratio. Technophobes may be put off.
Dall-E 3
OpenAI’s Dall-E 3 is probably the best starting place in AI art for those willing to pay for the service, as it is accessed through the highly intuitive ChatGPT chat bot.
Access is granted as part of a ChatGPT Plus subscription, which costs $20 a month. It also unlocks access to a more advanced version of the ChatGPT chat bot.
Comparing like-for-like, Dall-E 3 will often produce more cartoony-looking images than Midjourney v6. But this will naturally vary between prompt, and often its style may be just what creators are looking for.
Photoshop AI Generative Fill
Adobe added AI Generative Fill to Photoshop in 2023. As such, it instantly became one of the most important AI art tools in existence.
It’s a generational evolution of a feature Photoshop already had, the content-aware fill. But where that old version would typically be used to expand relatively plain backgrounds in images, AI Generative Fill can be used to enlarge pictures, making up extra image information.
There’s a free trial of Photoshop available, but longer-term access required an Adobe Creative Cloud subscription. These start at $19.99 a month for plans that include Photoshop.
Stable Diffusion
Stablity AI’s Stable Diffusion was one of the original big three names in generative AI imagery, alongside Dall-E and Midjourney.
It’s particularly interesting because Stable Diffusion is free and open source. You can’t access it directly from the Stability AI website, as a membership is required to use the servers that actually do the work.
However, there are other services that use their own servers, and can be used in limited fashion for free. Try out Stable Diffusion for free over at Playground.
Leonardo AI
Leonardo is one of the more friendly and useful interfaces for creating AI art. While there are paid subscriptions for this one, you also get a limited number of tokens to create custom images each day.
It’s at least partly based on Stability AI’s core intelligence, so the results may be similar to what you’d see on Stable Diffusion. However, the range of tools it puts at your fingers is impressive, including transformation of rough sketches and creation of video.
How Is AI Changing Art?
The most roundly positive take on AI art is that it democratises art creation. If you have a newsletter, for example, you could use it to make a header image. Or those who want to try making their own homespun indie video games could use AI to hide a lack of natural artistic ability (or time to practise and make up for that).
These AI image tools can also be used to dramatically improve the look of websites or presentations, without reverting to the clip art horrors the older folks among us may remember using.
However, this still ends up devaluing the work of artists, and further normalises the idea there’s no reason for people to be paid to create artwork. This isn’t just about people making and selling paintings either. In fact, it’s hardly about that at all.
Storyboard artists for film and TV, illustrators for books and advertising, videographers, commercial photographers and designers of corporate logos can and will, to at least some extent, be replaced by AI. This is particularly galling for those working in such industries when image-creation AI is trained in part using imagery they have created.
Where they are not replaced, artists may end up as editors and curators of content created by AI. This could turn traditionally made digital and physical art into a heritage industry, and it will greatly reduce the number of “normal” jobs involved in these pursuits.
What the jobs outlook will seem like at that point is as yet unknown. OpenAI boss Sam Altman has no worries about AI eradication of jobs (his net worth is measured in the hundreds of millions of dollars, after all), saying “we’ll find new and better jobs.” But for most those new jobs may not include the kind of art-making some have aspired to since early childhood.
What’s up next? AI video generation, as demonstrated by OpenAI’s Sora tool. This works much like image generation tools, working off a text prompt. Film director Tyler Perry claims he indefinitely suspended an $800 million expansion of his movie studio after seeing how impressive Sora is.
AI video generation isn’t new, but before Sora it tended to produce fairly nightmarish-looking footage.
Bottom Line
AI art is a powerful and precarious topic. While artificial intelligence is due to affect most jobs, it does so very visibly in AI art. We can all play around with image-generating AI, and understand intuitively that before its existence it would have taken hours (maybe hundreds of hours in some cases) to achieve the same results by hand.
Using AI art tools is a useful way to get your head around how these super-popular forms of AI work, though. You tend to have to pay to access the best of them, but if your interest is based around pure curiosity there are free alternatives available online. And countless YouTube videos demonstrating the power of the latest models are free to watch.
