Our team tests, rates, and reviews more than 1,500 products each year to help you make better buying decisions and get more from technology.
AI image generation is now fully mainstream, but that doesn’t mean you will get perfect results every time. Even the best services we’ve tested occasionally produce images with errors and distortion. We’ve reviewed AI image generators since their inception and subjected each to a series of diverse tests to inform our findings. Google’s Gemini and OpenAI’s ChatGPT are our current favorites thanks to the accuracy and detail of their images, but each of the services below is worth exploring. Make sure to scroll past the list for answers to all your pressing questions about generating AI images.
Adobe Firefly offers more than just Adobe’s AI technology. Yes, you can generate images with Firefly’s in-house models, but you can also use those from ChatGPT, Gemini, and a variety of others. Firefly’s models allow you to customize your generations in numerous ways, including aspect ratio, composition, effects, style, and more, which is a unique feature among AI image services. The premium version of Firefly also costs just $10 per month, whereas many other options start at $20 per month.
AI newbies: Not sure which AI image generation model will be most to your liking? Firefly helps you explore a wide variety of choices.
Visionaries: If you have a specific vision in mind, Firefly provides a plethora of intuitive ways to adjust the specific aspects of images you generate. For such specific requests, however, it is best to stick with Adobe’s in-house model.
Although ChatGPT has lost its title as the best AI image generator overall, it still impresses with sharp images that rarely have major issues. It also generally does a good job with complex prompts, such as technical diagrams, compared with the competition. For editing, you can use Adobe apps inside ChatGPT to make changes to your photos. You also gain free access to ChatGPT’s latest image-generation technology, though with limitations.
Adobe fans: You can leverage Adobe Express and Photoshop in ChatGPT to make localized brightness, contrast, saturation, and other adjustments via prompts and sliders. That integration is helpful if you like how Adobe’s tools work.
Free users: If you want to generate AI images for free, ChatGPT merely limits the number of generations you can create and the speed at which you can generate them, rather than forcing you to use a less advanced model.
Gemini’s Nano Banana Pro image generation model is best-in-class, offering excellent fidelity across the board. Whether you want to combine multiple images, generate text, or even tell a cohesive story through comic panels, Gemini excels in areas in which others often struggle. We like that you can use Nano Banana Pro for free, albeit with limits. And once you exhaust your allotment, you can still use the original (and still quite capable) Nano Banana model.
Image editors: If you’re looking for an easy way to edit images with AI, Gemini is your best option, thanks to the quality and realism of its edits. It’s simple to prompt it to do what you want.
Pixel-peepers: Most AI image generators restrict you to relatively low-resolution outputs, but Gemini can output high-resolution AI images in 2K or even 4K. Just note that it might still introduce blurriness to the final result.
Grok is unique among the AI image generators we tested in that it allows you to create content with explicit elements, including nudity. Although you should avoid doing so for ethical and legal reasons, Grok allows you to generate explicit content of real people, too. This functionality aligns with Grok’s so-called uncensored approach to AI technology, as well as its other unique features, such as its NSFW anime character companions.
NSFW creators: If you want to generate NSFW AI images easily and without any fuss, Grok is the way to go. It can also turn your generations into videos.
X users: Grok is a good fit if you often use X, as it can search that platform in response to your queries.
Copilot can generate images just like ChatGPT or Gemini, but you can also leverage it from within Microsoft’s Paint app on Windows 11. Aside from generating images based on prompts, it’s also possible to remove backgrounds and other objects using AI. These features are free, as is Copilot in general, but you need to pay for a Microsoft 365 plan to increase your usage limits
Windows users: Paint is exclusive to Windows PCs, so macOS and mobile users need to look elsewhere for similar functionality. Copilot also features several Windows-exclusive capabilities on Copilot+ PCs, including Click to Do and Recall.
Although you might know it best for its stock photo catalog, Shutterstock’s AI image generator stands out by offering a large variety of AI image generation models. You can use Shutterstock’s proprietary model or those from Amazon, ChatGPT, Gemini, and more. Alternatively, you can let Shutterstock automatically choose the best option for the task at hand. Each Shutterstock generation gives you four images to choose from, which massively cuts down on how many generations you need to do for each prompt you submit. Moreover, you get to create 100 generations (400 unique images) for just $15 per month, billed annually.
Model-agnostic creators: If you don’t have a favorite image-generation model, Shutterstock makes it easy to sample from a wide variety of them. By generating multiple images per request, you can gain a good sense of how well each one meets your needs.
Nostalgists: Among Shutterstock’s models are classic ones that may not be easily accessible elsewhere, such as OpenAI’s DALL-E 3 or Gemini’s older Imagen 3 and Imagen 4 models.
Evaluating how each AI image generator handles our test prompts is the primary way we determine which services make this list. We consider only full-service AI image generation services, meaning those that can be accessed or downloaded directly on the web. It’s possible, of course, to generate AI images locally on your computer, but these solutions require quite a bit of setup time and processing power.
Our testing is tailored to each service’s unique features, but our evaluations generally encompass a few key categories: basic scenes, complex actions, and text. We vary our prompts between rounds of testing since AI images can include so many different elements and mimic various styles, but below are the basic frameworks:
- Basic scenes: We request a relatively simple depiction of the interior of a home, featuring a diverse array of objects. This test lets us evaluate a product’s ability to generate photorealistic images.
- Complex action: We prompt the services to create a comic with multiple panels that tell a coherent story throughout, culminating in a twist by the final panel.
- Text: We task the image generators with producing a diagram that shows how to set something up (much like an instruction manual), along with labels for the various items and steps involved.
As for image editing, we focus on a service’s ability to localize edits to a specific area rather than generating an entirely new image from the image and prompt. Not all AI image generators specialize in AI image editing, though, so don’t expect a service with high marks on one to necessarily do as well with the other.
Which AI Image Generators Didn’t Make the Cut?
Popular AI image generation models, such as Midjourney and Stable Diffusion, don’t make this list. They can perform well, just not consistently at the level of top models. Furthermore, they are technically closer to models than services. For example, you can use Midjourney tech on Meta’s site and Stable Diffusion with ComfyUI. Finally, these Midjourney and Stable Diffusion are better at artistic generations than realistic ones, which may or may not suit your purposes.
How Well Does AI Image Generation Work?
AI image generation can be very impressive, but you likely need to do multiple generations, pay for a higher-tier service, and tweak your prompts to achieve the best output quality. After all, minor changes and optimizations can lead to drastic differences. That said, AI image generation tech continues to improve rapidly, so we expect the process to become increasingly affordable and simple.
Is AI Image Generation Free?
In many cases, you can create AI images for free. However, you should expect restrictions on the number of images you can generate and limited access to the most advanced models. For example, you can use Gemini’s Nano Banana Pro model for free until you hit a cap, after which Google downgrades you to the original Nano Banana model (which is still good enough to earn our Technical Excellence Award).
How Can You Tell If an Image Is Real?
The short answer is that you can’t always tell. The longer answer is that a variety of signs can indicate that an image is likely to be fake. However, as AI images become more realistic, these signs might not be so obvious.
Can You Generate AI Images of Yourself or Other Real People?
Yes, but different services place various restrictions on this functionality. For example, Copilot refuses to generate images of real people, but Grok is happy to do so. Just don’t be the person who makes fake images to harass people or spread misinformation; nobody likes that person.
Can You Make NSFW AI Images?
You absolutely can, and on this list, Grok is your best bet for that, considering most services don’t allow it. Still, you are likely better off finding a dedicated service that specializes in NSFW AI image generation. Check out our list of the best NSFW AI image generators for the full story.
Disclosure: Ziff Davis, PCMag’s parent company, filed a lawsuit against OpenAI in April 2025, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.
Read Our Editorial Mission Statement and Testing Methodologies.
My Experience
I’ve been writing about consumer technology and video games for over a decade at a variety of publications, including Destructoid, GamesRadar+, Lifewire, PCGamesN, Trusted Reviews, and What Hi-Fi?, among many others. At PCMag, I review AI and productivity software—everything from chatbots to to-do list apps. In my free time, I’m likely cooking something, playing a game, or tinkering with my computer.
The Technology I Use
I use a ThinkPad for work, but my heart belongs to the PC I built with a fully custom water-cooling loop down to the SSD. Outside of that, I usually hang onto a Pro Max iPhone for a couple of years before getting the latest model. I also spend a decent amount of time with an aging Kindle.
As for software, I’ve used Chrome and iTunes for too long to stop. I rely on the Google Suite for organization and backing up my data, and I couldn’t enjoy my days off without Discord and Steam. I typically write down what I need to do in the Notes app on my iPhone.
For audio, I’m a lover of cables, especially the ones that connect to my Shure SRH-1540 daily drivers. At home, my Yamaha RX-V583 receiver drives a pair of Paradigm Prestige 15Bs for stereo entertainment, with enough Polk speakers in concert to round out a 7.1 setup.
I’ve been writing about consumer technology and video games for over a decade at a variety of publications, including Destructoid, GamesRadar+, Lifewire, PCGamesN, Trusted Reviews, and What Hi-Fi?, among many others. At PCMag, I review AI and productivity software—everything from chatbots to to-do list apps. In my free time, I’m likely cooking something, playing a game, or tinkering with my computer.
I use a ThinkPad for work, but my heart belongs to the PC I built with a fully custom water-cooling loop down to the SSD. Outside of that, I usually hang onto a Pro Max iPhone for a couple of years before getting the latest model. I also spend a decent amount of time with an aging Kindle.
As for software, I’ve used Chrome and iTunes for too long to stop. I rely on the Google Suite for organization and backing up my data, and I couldn’t enjoy my days off without Discord and Steam. I typically write down what I need to do in the Notes app on my iPhone.
For audio, I’m a lover of cables, especially the ones that connect to my Shure SRH-1540 daily drivers. At home, my Yamaha RX-V583 receiver drives a pair of Paradigm Prestige 15Bs for stereo entertainment, with enough Polk speakers in concert to round out a 7.1 setup.
Read full bio
PCMag.com is a leading authority on technology, delivering lab-based, independent reviews of the latest products and services. Our expert industry analysis and practical solutions help you make better buying decisions and get more from technology.
For over 40 years, PCMag has been a trusted authority on technology, delivering independent, labs-based reviews of the latest products and services. With expert analysis and practical solutions across consumer electronics, software, security, and more, PCMag helps consumers make informed buying decisions and get the most from their tech. From in-depth reviews to the latest news and how-to guides, PCMag is the go-to source for staying ahead in the digital world.
Leave a Reply