AI image generators have been brewing (generating?) up a storm for more than a year. If you’ve been on social media, watched prime time news shows, or read a magazine, AI-generated images have been impossible to miss. They’re everywhere, and it’s easy to see why: the tools necessary to make them are now good—and available to the public. If you want to join in the fun, or add some AI-powered features to your business workflows, the apps on this list will give you what you’re looking for.
I’ve been writing about AI image generators since Google Deep Dream in 2015. That’s about as long as anyone realistically has been thinking about these tools, and it’s really exciting for me to see how far they’ve come.
I’m going to try to avoid the thorny discussions around artistic merit, whether or not these tools are replacing or augmenting artists, and copyright infringement in training data, at least where I can. Instead, I’ll focus on the fact that these AI image generators can now produce fascinating results from written prompts. It’s worth taking a few hours to play around with one of these text-to-image AI apps—even just so you can appreciate them from a technical perspective.
The best AI image generators
How do AI image generators work?
All these AI image generators take a text prompt and then turn it—as best they can—into a matching image. This opens up some wild possibilities, since your prompt can be anything from “an impressionist oil painting of a Canadian man riding a moose through a forest of maple trees” to “a painting in the style of Vermeer of a large fluffy Irish wolfhound enjoying a pint of beer in a traditional pub” or “a photograph of a donkey on the moon.”
Seriously, the only real limits are your imagination, the AI image generator’s ability to comprehend your prompt, and any content filters put in place to stop bad actors flooding the internet with AI-generated violence or other NSFW content.
Most AI image generators work in a pretty similar way. Billions of image-text pairs are used to train a neural network (basically, a very fancy computer algorithm modeled loosely on the human brain) on what things are. By allowing it to process near-countless images, it learns what dogs, the color red, Vermeers, and everything else are. Once this is done, you have an AI that can interpret almost any prompt—though there is a skill in setting things up so it can do so accurately.
The next step is to actually render the AI-generated image. The latest generation of AI image generators do that using a process called diffusion. In essence, they start with a random field of noise and then edit it in a series of steps to match their interpretation of the prompt. It’s kind of like looking up at a cloudy sky, finding a cloud that looks kind of like a dog, and then being able to snap your fingers to keep making it more and more dog-like.
Before we dive in: I don’t want to oversell things. What these text-to-image generators can do is very impressive, but they aren’t likely to save you from ever having to do a product photoshoot again. If you just need some weird or unique images, they can really help. But if you’re looking for something super specific, you’re better off hiring a photographer—or licensing the exact image you want. Similarly, trying to use one to make a header image for a blog post can take a lot more time than just finding a header image for your blog through a stock photo site. Sure, it won’t be as custom, but the model is more likely to have ten fingers.
What makes the best AI image generator?
How we evaluate and test apps
All of our best apps roundups are written by humans who’ve spent much of their careers using, testing, and writing about software. We spend dozens of hours researching and testing apps, using each app as it’s intended to be used and evaluating it against the criteria we set for the category. We’re never paid for placement in our articles from any app or for links to any site—we value the trust readers put in us to offer authentic evaluations of the categories and apps we review. For more details on our process, read the full rundown of how we select apps to feature on the Zapier blog.
There’s a reason that AI image generators have become incredibly popular over the past 18 months: before that, they were pretty bad. The technology underlying them was incredibly cool and impressive, at least to research scientists, but the images they could output were underwhelming. Even the original DALL·E was more of a fun novelty than a world-shaking revelation when it launched in 2021.
Since they’re such a recent development, there isn’t a huge amount of great AI image creators to choose from. The three biggest names—DALL·E 3, Midjourney, and Stable Diffusion—are all on this list for different reasons, and you’ll also see two apps built by the large enterprises that these tools were initially disrupting—in this case, Adobe and Getty. But also, these five apps were really the only ones that met my criteria for inclusion for two reasons:
I was looking for apps that allowed you to generate AI images from a text prompt. Tools that have you upload a dozen of your photos and then spit out AI-generated portraits are fun (and normally built using Stable Diffusion), but they aren’t the kind of general-purpose image generators I was considering.
I was looking at the AI image generators themselves, not tools built on top of them. For example, NightCafe is an AI picture generator that has a great community and app, but it just enables you to use the DALL·E 2 and Stable Diffusion algorithms (as well as some of the older ones). It’s worth checking out, but it doesn’t meet my criteria for this list.
Aside from all that, I also considered how easy each AI image creator is to use, what kinds of controls and customization options it provides (for things like AI image upscale), what pricing model it has, and most important of all: how good were the results?
I’ve been using and writing about DALL·E, Midjourney, and Stable Diffusion since they all launched, and about Photoshop and Getty Images for more than a decade, so I’m pretty familiar with how they all work—and their various pros, cons, and weird behaviors. But writing this article was actually the first time I’ve put all five head-to-head with the same prompts. The results were fascinating, and I’m delighted to say all five offer genuine reasons to use them.
Before we dive in, one more thing to note: all these tools are technically in beta, and I suspect they’ll remain that way for a while. While they’re getting more and more impressive every day, AI image generators have a long way to go before they’re able to consistently produce great results and reliably fit into commercial workflows.
Note: Unfortunately, being in beta doesn’t mean these tools are free. Since I first wrote this article, both DALL·E and Midjourney have scrapped their free trials. It’s understandable, given the heavy computing load involved in creating AI images, but it still means that they’re harder to just play around with.
How to use AI image generation at work
Interested in AI, but not quite sure how you’d use it at work? Here are a few of the ways people are turning to AI image generation in their roles:
Generating hero images for blog posts
Creating social media posts
Generating slide decks and storyboards
Creating personalized images for customers
Learn more about how to use AI image generation at work.
The best AI image generators at a glance
Ease of use
Included with ChatGPT Plus at $20/month
From $10/month for ~200 images/month and commercial usage rights
Midjourney (but it’s accessed via Discord)
Customization and control
Free for 25 credits; from $10 for 1,000 credits
Integrating AI-generated images into photos
Free beta online; from $19.99/month as part of the Creative Cloud Photography Plan
Commercially safe images
You need to contact them for a demo and pricing
The best AI image generator for ease of use
DALL·E 3 pros: Incredibly easy to use; included with ChatGPT Plus, so you get a lot of AI for your money
DALL·E 3 cons: ChatGPT controls can be hit and miss; no longer has a free trial; $20/month is pricey if you don’t want GPT with it
DALL·E 3 is arguably the biggest name in AI image generators—and with good reason. Its predecessor, DALL·E 2, was the first AI-powered image generator that was good enough to create wildly interesting images and was widely available to enough people to go viral.
And while DALL·E 2 is still available through its web app and API, DALL·E 3, which you can use through ChatGPT, is a significant improvement. For any given prompt, it produces more interesting, more realistic, and more consistent results. Previously, it felt like OpenAI was falling behind its competitors with AI image generators, but DALL·E 3 has brought it right back into the race.
The biggest thing is that DALL·E 3 is ridiculously simple to use. Tell ChatGPT what you want to see, and within a few moments, you’ll have four AI-generated variations to choose from. It uses GPT’s understanding of language to expand your prompts, so each result will be distinct.
Unfortunately, DALL·E is no longer free to try out. If you’re a ChatGPT Plus subscriber, you can use it as much as you like, subject to GPT-4’s 50 messages every three hours limit. Or, if you’re happy to stick with DALL·E 2, you can get 115 credits for $15—that’s around $0.13 per prompt or $0.0325 per image variation. While it’s a lot more basic, at that price, it’s still worth considering.
Right now, DALL·E 3 doesn’t have all the same features as DALL·E 2. You can ask ChatGPT to make changes, but you can’t do too much otherwise. Sometimes this feels like magic, where ChatGPT will do exactly what you request. Other times, it’s like working with an overeager intern who’s a touch too happy to do their own thing.
While DALL·E 2 produces more basic results, it still has some powerful features. The image editor (which is also in beta) enables you to add additional generated frames, so you can expand an image, whether you generated it using DALL·E 2 or uploaded it. This technique, called out-painting, means you can create larger works of AI art. There’s also an eraser, so you can remove bits of an image and replace them with AI-generated elements (this technique is called in-painting). These features don’t perfectly map to DALL·E 3 yet, though you can ask ChatGPT to expand your image or remove things you don’t like, which can give you a similar effect.
In addition to DALL·E 3 through ChatGPT and the DALL·E 2 web app, OpenAI offers an API, which allows developers to build apps that integrate with the two DALL·E models (though support for DALL·E 3 is still forthcoming). Because of that, you can connect DALL·E to Zapier to do things like automatically create images from Google Forms or HubSpot responses—or any other apps you use. Or you can use it to automate your art inspiration.
Zapier is a no-code automation tool that lets you connect your apps into automated workflows, so that every person and every business can move forward at growth speed. Learn more about how it works.
DALL·E pricing: DALL·E 3 is included as part of ChatGPT Plus at $20/month; DALL·E 2 costs $15 for 115 image credits. API pricing is more complex, but starts from $0.016/image.
The AI image generator with the best results
Midjourney pros: Consistently produces the best looking AI generated images; the community is a great way to get inspiration
Midjourney cons: Can only be used through Discord—which is just strange; images you generate are public by default; free trials are currently suspended
Midjourney consistently produces my favorite results of all of the image generators on this list. The images it creates seem more coherent, with better textures and colors—and overall, the results are just more interesting and visually appealing. In particular, people and real-world objects look more lifelike and natural than they do with other AI image generators, at least without lots of prompting, and the latest version can even get hands kind-of right. It’s telling that it was the first AI image generator to win an art competition.
Unfortunately, Midjourney is let down by one major quirk: the beta is only accessible through Discord. Once you join Midjourney’s Discord server or invite the Midjourney bot to one you control, you can enter a prompt by typing
/imagine [whatever you want to see]. The bot will then generate four variations of your prompt, which you can then download, upscale, re-edit, and more.
By default, every image you generate is posted publicly in Midjourney’s Discord. It gives everything a cool community aspect, but it means that anyone who cares to look can see what you’re creating. While not necessarily a problem for artists, this might be a dealbreaker if you’re looking to use Midjourney for business purposes.
If all this sounds confusing, don’t worry. Midjourney’s help docs are really good and walk you through getting started as well as all its advanced features, like the different model versions, upscaling your images, blending multiple images, and using different parameters to control things. Once you understand the different options, the results you can get are genuinely amazing.
Midjourney’s free trials are currently suspended because of the overwhelming number of people trying to use it, but they’re occasionally reinstated for a few days. If you miss a free trial window, the Basic Plan starts at $10/month and comes with 3.3 hours of GPU time per month, or around 200 images. You also get the option to buy additional GPU time, and you can use your images commercially.
Midjourney pricing: From $10/month for the Basic Plan that allows you to generate ~200 images/month and provides commercial usage rights.
Best AI image generator for customization and control
DreamStudio (Stable Diffusion)
DreamStudio pros: Only major AI picture generator that still offers free credits; incredibly affordable and customizable; super powerful with generally great results
DreamStudio cons: More of a learning curve than some other AI art generators; editing tools aren’t very intuitive
Unlike DALL·E and Midjourney, Stable Diffusion is open source. This means anyone with the requisite technical skills can download it and run it locally on their own computer. It also means that you can train and fine-tune the model for specific purposes. Almost all the services that use AI to generate artistic portraits, historical portraits, architectural renders, and everything else use Stable Diffusion this way. If you’ve got the chops and want to build something awesome with AI, Stable Diffusion is the best way to do it right now. There’s even an API and a service called ClipDrop that breaks it out into a series of tools.
DreamStudio gives you a huge amount of control over the various aspects of generating an image with AI. When you type in your prompt, there are sliders that allow you to determine how large the final image is, how closely it matches the prompt you give it, how many steps the diffusion model takes, and how many images are generated. You can also select what version of the algorithm it uses (the latest is SDXL 1.0), and even enter a specific seed so that you get repeatable results (otherwise, they’re randomly generated). DreamStudio also has in-painting and out-painting, though you need to use Chrome to access them, and more editing features are apparently coming soon.
DreamStudio works on a credit system. When you sign up, you get 25 free credits, which are good for around 30 prompts or 120 images with the default settings. Using a more powerful model, generating larger or more images, or iterating them through more steps will all use up your credits faster. Once you’re done, you’ll need to buy more, starting at $10 for 1,000 credits.
All in all, DreamStudio and Stable Diffusion give you the most customization and control over the whole AI image generation process. They enable you to go as deep into AI as you want to go—and even build your own AI services.
And if you burn through your free trial too quickly, you can also try the same Stable Diffusion models for free through ClipDrop—though they’ll be watermarked, and you have less control.
DreamStudio pricing: Free for 25 credits. From $10 for 1,000 credits.
Best AI image generator for integrating AI-generated images into photos
While Adobe has been building AI tools into its apps for more than 15 years, it wasn’t until this year that the company released a text-to-image generator—at least in beta. You can try it out on the web for free or through Adobe Express, but it’s at its best in the latest Photoshop beta, which you need to be a Creative Cloud subscriber to check out.
Adobe’s AI model, called Firefly, has a few tricks up its sleeve. In addition to being capable of generating new images from a detailed text description, it can create text effects from a written prompt (think, the word “TOAST” written with letters that look like they’re made from toast), recolor vector artwork, or add AI-generated elements to your images. You can test all these out through the web app, but it’s that last feature where Firefly stands out.
Taken purely as a text-to-image generator, Firefly’s results can be pretty hit and miss. It can match or beat DALL·E or Stable Diffusion for some prompts, but for others, I question what it was aiming to do. On the other hand, its integration with Photoshop, the industry standard image editor, is next level.
The particular feature is called Generative Fill. The idea is that you use Photoshop’s regular tools to select an area of your image, and then, just by clicking a button and typing a prompt, you can replace it with something else. Crucially, Generative Fill understands the context of your image. In the screenshot above, you can see that Photoshop has matched the depth-of-field blur and colors for the castle I added using Generative Fill. It looks cohesive.
As much as DALL·E and Stable Diffusion have started the conversation about image-generating AIs, Adobe’s Firefly is the first implementation of an AI photo generator that really hints at what’s to come. It isn’t a party trick, but a tool that will soon be available to the millions of professionals who use Adobe apps every day.
It’s worth noting that while Firefly is in beta, the images it generates aren’t supposed to be used for commercial purposes.
Firefly pricing: Free beta online; Photoshop is available from $19.99/month as part of the Creative Cloud Photography Plan.
Best AI image generator for usable, commercially safe images
Less creative and fun to use
Can’t compete with Midjourney, DALL·E 3, or Stable Diffusion in terms of overall quality
As we’ve discussed, AI image generators are… controversial. Even aside from the ethics of using them, the legal situation is incredibly murky. The U.S. Copyright Office has generally ruled that AI images can’t be copyrighted, so your competitors could conceivably take your images and use them without repercussions. Honestly, if you run a business, just avoiding generative AIs is an understandable strategy. But if you really want to use them, then a platform like Getty Images that promises its Generative AI image generator is free from these sorts of headaches is the best option.
Despite the awkward title, Generative AI by Getty Images is surprisingly good—especially at creating stock-like photos. I tested it with classics like “woman laughing alone with salad,” and the results were pretty solid. At a glance, the best options were indistinguishable from real stock photos.
For more creative prompts, like “a Canadian man riding a moose through a maple forest,” and anything involving specific art styles, it was a bit rougher. I suspect this is down to the training data. Instead of a questionably-sourced collection of a few billion images, Generative AI was largely trained on Getty’s stock image catalog. According to Getty, this was all done above board, and it has a program in place to compensate artists whose work was used to train the model.
While that’s laudable, it does affect the kind of things you can generate. Generative AI won’t create anything featuring a real person, a trademark, or anything else that potentially violates intellectual property laws. I couldn’t even get it to generate a painting in the style of Vermeer, despite the fact he died in 1675. Overall, it makes Getty’s tool less fun and flexible to use—but much more practical for companies with a legal department.
Generative AI by Getty Pricing: Contact Getty for a demo and pricing
Stock image site Shutterstock also has an AI image generator that it built with OpenAI. I tested it out and didn’t like it as much, but it’s worth checking out if you need commercially safe images.
How to use an AI image generator
Ok, so you know what the best options are, but…now what? The team at Zapier has put together a bunch of resources to help you understand how to use these tools—and put them to work.
First, tutorials and walkthroughs for each of the best AI image generators on this list:
Plus, a guide for how to write effective AI art prompts, so you can get what you’re looking for faster (and better) when generating images.
Once you’ve got the basics down, it’s time to use these tools for more than just creating wacky pictures. Here are some tips for how to use AI image generators at work:
And finally, a few ways that you can automate your AI image generators, so they do their magic behind the scenes and connect to all the other apps you use.
What about all the other AI image generators?
You probably noticed that this list is pretty short—I only picked five AI image generators. As I mentioned above, that’s because I’m looking at the AI image models themselves—not necessarily the apps that are built on top of them.
If you want a laundry list of every AI image generator out there, including those that are built on top of these three models, I made that too. It includes two dozen image generators: some are built into other tools, like AI writing apps, photo editing apps, or stock photo sites; some let you select from multiple models; and each one differs on how it approaches AI image generation. So if none of the apps on this list feel natural to you, check out my list of the top AI art generators, and see if anything stands out.
The legal and ethical implications of AI-generated images
AI-generated images are everywhere now, but that doesn’t mean we shouldn’t be asking questions about how they should (or shouldn’t) be used.
There aren’t clear laws in place surrounding AI-generated images. And that goes for both sides of the coin: the U.S. Copyright Office suggests that AI-generated content isn’t copyright-protected, and there aren’t rules to protect artists whose work was scraped for AI training. (That’s why Firefly was trained on licensed images and public domain content only.)
It’s already led to lawsuits. Stability AI, for example, is facing lawsuits from Getty Images and artists themselves for unauthorized use of their images, and there’s a class action lawsuit against a number of AI picture generators.
You’re not likely to get into trouble for using AI-generated images for a few social media posts or blog hero images, but because there’s no line drawn in the sand yet, it can be risky to develop an entire strategy around AI-generated art.
Then there’s the issue of bias. As of now, AI has a lot of the same biases as humans, and that can lead to everything from the portrayal of stereotypes to harmful content. I experienced this myself with the outputs I got while testing these apps. It’s up to us as humans to avoid it by reviewing AI-generated content for bias and refining our prompts to eliminate that bias as much as possible.
What’s next for AI image generators?
AI image generating is a rapidly evolving space. It’s incredible to see how far the different engines have come over the space of a year. With hundreds of thousands of people now using them, the developers are getting huge amounts of data to train and refine their models more, so we can expect things to continue to improve.
It’s also likely that we’ll soon see some new image generators get released. Google hasn’t yet made Imagen publicly available, and Meta hasn’t released anything based on its Make-A-Scene algorithms to the public—exciting times are ahead.
This article was originally published in March 2023. The most recent review of apps and update was in October 2023.