Fueled by the rise of MySpace and a secret Bollywood fandom, I started teaching myself graphic design as a preteen. But thanks to AI image generators, you no longer have to be a lifelong creative like me to turn your ideas into a visual reality.
AI image generators can help anyone create visuals that are weird, wonderful, and…weirdly wonderful. (Remember the Willy Wonka debacle in Glasgow?)
But how exactly do you use AI to generate images, and what’s the best option for marketers? I tested the top options — Here’s the scoop.
Table of Contents
How do AI image generators work?
AI image generation can get wild. You can create images in any style and of practically anything. For instance, I made “a photorealistic orange rabbit wearing a traditional Indian sari and playing an acoustic guitar” using Google’s Gemini.
Then I asked for an electric guitar:
Why exactly? Well, as many people justify their oddball AI creations: because I could. But that doesn’t mean using AI to generate images should be taken lightly.
Plagiarism is one of many concerns with even the best AI image generators. And it makes sense, considering the very nature of how they work. AI image generators create by reimagining things that already exist. They don’t copy. They borrow and recycle.
- The AI is “trained” on millions of existing images, descriptions, and captions available across the internet. This is how it “learns” what things look like and what they are called.
- When you enter a prompt into a tool, the AI reaches into its memory of those images and captions.
- The AI produces an image or selection of images it believes fit your prompt.
Think of it like a contemporary singer sampling a song from a bygone era. They’ve made it their own, but pieces of the original are still there and recognizable.
How I Tested the Best AI for Images
For the purpose of this article, I focused on a common marketing use case: creating an original image for a blog article.
I tested nine of the most popular AI image generators and evaluated them on their speed, ease of use, and image quality.
To fairly compare the options, I also eliminated variables by:
- Using the same prompt
- Requesting the same edit to the original prompt
I also ran each tool three times after each prompt, giving them a fair opportunity to deliver. Consider these my “trials.” This is science, after all.
The prompt: “Generate an image of a young marketing executive presenting information on a television screen in front of a small group of teammates.”
The edit: Generate a photo-realistic image of a young marketing executive presenting information on a television screen in front of a small group of teammates.
What makes an AI image generator good?
To answer this question, here are the factors I evaluated each tool on:
- Ease of Use: How easy it is to prompt the tool and get results
- Speed: How fast it took to generate the images,
- Image Quality: How high resolution the image is; its clarity, accuracy, etc.
- Diversity and Variety: How original the selection of images is as well as how diverse and inclusive they are with human depictions.
- Customization Options: How specific you can get with your prompt and editing features.
If a tool is fast, easy to use, customizable, and delivers high-quality, creative images, it’s top tier. With all this in mind, let’s get to our test results.
1. HubSpot AI Image Generator
HubSpot’s AI image generator is currently in beta, but it is one of many AI-powered features rolling out across our tools. It can be used within your HubSpot portal to create images for emails, blogs, landing pages, or repurposing with Content Remix.
Pricing: Starting at $450/mo with Content Hub, depending on your plan.
Test Results
HubSpot’s AI image generator can be found across a number of tools, but for the purposes of this experiment, I navigated to the blog. From here, I hit “Insert image” then clicked on the “Generate with AI” button.
I was met with three simple fields to fill out (seen below):
- “Describe your image,” where I would enter my test prompt
- “Choose a style,” where I could pick from 17 different options or allow the tool to decide.
- “Dimensions” where I could choose between square, landscape or portrait.
I entered my details, hit “generate,” and was met with two images in about 10-20 seconds. I was impressed by the speed! Then, I could like or dislike each image, crop and resize them within HubSpot, or edit them using Adobe Express.
At a glance, the thumbnails look pretty promising, but, honestly, when I enlarged them using the “crop and resize” button, I could see there were some very noticeable issues with hands and facial features.
This is common with many AI image generators, so it wasn’t a big surprise.
The images also were both of individuals who presented as White males with rooms filled with White team members.
For my second trial, I used my prompt specifying “photo-realistic” but let the tool decide the style.
The results were actually less photo-realistic, in my opinion, and the feature issues continued, but I was happy to see more diversity in the second iteration.
With its speed and ease of access, HubSpot’s built-in AI image generator is a solid solution for users, but as it's in beta, there is room to grow. I’d love to see the issues with fine details like facial features refined and more diversity in human depictions.
All that said, the images are still great for use in smaller sizes.
What I like:
- Fast, easy execution
- A lot of style options
- Easily accessible within the HubSpot tools; no need to jump platforms
What can be improved:
- Only two results
Best for: HubSpot users who need simple, small images to break up their content
2. Midjourney
Midjourney is considered one of the most powerful generative AI tools out there, so my expectations for its image generator were high. It focuses on creating artistic and stylized images and is popular for its high quality.
Pricing: $10-120 per month, depending on the plan
Test Results
I’m used to the chatbot setup of ChatGPT, so Midjourney wasn’t what I expected.
To create, you have to join the Midjourney Discord channel (similar to Slack). From there, you use keyboard commands within chats to have the Midjourney bot perform your desired tasks.
In this instance, I had to join a “newbie” chat (there were several to choose from) and enter the command “/imagine” followed by my prompt. Then, the Midjourney bot processed my request (and sometimes those of others simultaneously.)
It delivered four images with each trial.
From there, I could click numbered buttons underneath the images to get “upscales” (U) or variations (V) of a particular image. It isn’t entirely clear to a “newbie” what an upscale or variation means.
Personally, I wasn’t the biggest fan of this process. Anyone in the chat can see your prompt, the results, and even download them for their own use. Your results could also quickly be buried by others, and you’d have to scroll up to find them.
Speaking of results, they were quick but varied in quality.
The AI delivered a variety of styles and included some diversity in its human subjects (no glaring mistakes), but it produced the same settings and poses in each option.
Also, there was no clear way to edit the images created. I tried responding to the bot to no avail. My second prompt resulted in less photo-realistic images — even though I specified that’s what I wanted:
Overall, I was disappointed by Midjourney. It was frustrating to navigate and didn’t produce the quality I expected from the hype. Someone with more experience in Discord or prompting it might yield better results, but it didn’t win me over.
What I like:
- Fast (images in under 60 seconds)
- Multiple images at once
What can be improved:
- User-friendliness
- Privacy
- Editing capabilities
- Image style variety
- Photo-realistic quality
Best for: Artistic endeavors, marketers with advanced technical knowledge, web developers, or those comfortable within Discord who can learn to navigate the nuances of prompts.
3. DALL-E 3
Owned by OpenAI (the company behind ChatGPT), DALL-E is a pioneer in image generation.
DALL-E 3, the latest iteration of the tech, is touted as highly advanced and is known for generating detailed depictions of text descriptions. This means users can create original images and modify existing ones based on text prompts.
Pricing: It is available to ChatGPT Plus, Team, and Enterprise users, as well as developers through the API, starting at $20 per month.
Test Results
Unlike most tools on our list, DALL-E3 generated only one image at a time.
In a world where a search engine can find millions of pictures in seconds, this is highly limiting and, honestly, underwhelming. But the image quality was impressive.
It showed detailed, realistic portrayals of people and had no visible difficulty replicating fingers or eyes, as we often see with AI. It also allows users to edit specific image details:
For example, I requested the main subject of the image above shift to a woman of color and the information on the television screen change to an Instagram profile.
The photo after editing wasn’t as seamless. You could see where the AI spliced in the new content and certainly did not use an Instagram profile, but I digress.
As you may be able to guess from my requested edit, where DALL-E3 fell short was its lack of diversity and variety.
After running two more trials with my initial prompt, the settings of the image stayed the same — a generic, corporate office, while main subject in all of the images looked very similar as well, and most noticeably, all White or European.
While the results for the initial prompt were quite photo-realistic, I ran my second prompt.
The results were consistent, and the diversity issue also remained. Out of curiosity, I ran one more test in a new chat window and found that all images were now of men, but again, they all appeared to be White or European.
I was able to request changes to make the people in the image more racially diverse, but it took several tries.
Stereotyping and bias are common concerns with AI image generators, and that may be an issue with DALL-E3.
What I like:
- High-quality images (crisp, detailed)
- Realistic human depictions
- Ability to edit image details
What could be better:
- Only one photo per prompt
- Lack of diversity without additional prompting
- Lack of variety in the setting
- Lack of control over specifications (dimensions, style, etc.)
Best For: Brands that would benefit from ChatGPT's greater AI abilities. In my opinion, the cost is not worth it just for image generation, especially considering some of the amazing free options in this list.
Pro tip: If you find yourself getting the same results with each regeneration or requesting edits proves fruitless, try opening a new chat window.
4. Microsoft Designer
Designer is an AI image generator from Microsoft. It’s positioned as a tool to help you “create social media posts, invitations, digital postcards, graphics, and more, all in a flash.” Many say it’s a Canva competitor, and I can see why.
Designer uses DALL-E2 to generate images from text prompts, but you can also start with one of the built-in templates or tools.
Its user-friendly templates include stickers, collages, greeting cards, and social media posts. Users can also perform everyday editing tasks like removing a background from an image.
Pricing: Everyone gets 15 free generations per day, but this jumps to 100 with a Microsoft Copilot Pro membership for $20 per month.
Test Results
Coming from DALL-E3, I was immediately pleased to see Designer deliver four images with each run of my test prompt. I was also happy to see the diversity in gender, race, and even setting.
Like DALL-E3, the Designer results were realistic from the start (with no face or feature issues), but most still had an illustrative stroke.
Nevertheless, after running my second prompt, the results remained similar:
Once I found an image I liked, Microsoft Designer let me download, copy, edit, or use it in another design on the platform:
What I like:
- Multiple images at once
- High-quality images (crisp, detailed)
- Accurate depictions of people
- Lots of diversity (both in people featured and in style/setting of image)
What could be better:
- All images are square by default
- “Photo-realistic” images still have an illustrative look
Best for: Great option for anyone as a free tool, but especially valuable to teams looking for design flexibility or those heavily invested in social media. Microsoft Designer makes it easy to create graphics using its images.
5. Adobe Firefly
Firefly is design giant Adobe’s answer to generative AI. It can generate art or photo-style images in four common aspect ratios (square, portrait, landscape, and widescreen), and it allows users to select or upload resources for reference.
Pricing: Free for 25 credits per month, but premium plans offer more credits starting at $4.99 per month.
Test Results
Honestly, Firefly immediately blew me away.
Not only was it the fastest tool, but it also delivered four images in various styles, with a diverse group of subjects and some of the most photo-realistic results I’ve seen.
The depictions of humans were mostly realistic, but as I ran my additional trials, I did spot flaws like missing faces or choppy cut-outs in the backgrounds.
Since Designer has a built-in option for photos, I deviated a bit from my experiment. I ran the initial prompt under the art filter to evaluate the differences.
Again, they were striking. The results were a mix of vectors and illustrations that employed bold colors, precise details, and visual depth for beautiful results.
The platform also let me edit the images, generate more based on one I liked, and use any of the images in an Adobe Express design.
What I like:
- Free
- Fast
- Multiple images at once
- High-quality, photo-realistic images (crisp, detailed)
- Accurate depictions of people (mostly)
- Lots of diversity (both in people featured and in style/setting of image)
- Easy-to-use images in a design
- Edit specific details of images
What could be better:
- Have to inspect closely for flaws
- More image styles
Best for: Great option for anyone, especially as a free tool. Both the artistic and photo results are top-notch and easy to repurpose into design inside and outside of Adobe.
6. Canva Magic Design
Canva’s AI image generator, Magic Design, brings the power of AI to the masses. You can use it to generate images, graphics, or videos in square, vertical, or horizontal aspect ratios, and you can choose from over 20 visual styles.
Pricing: All users get 10 free Magic Designs over their lifetime, but if you want more, varying usage limits are available in Pro, non-profit, and Education accounts, starting at $120/year.
Test Results
With the initial prompt, Canva delivered four graphic/illustrated images in each trial. Many figures were simple vectors without any defining features, reminiscent of 1990s clip art.
You can see the results of each trial circled in the same color.
Upon entering my “photo-realistic” prompt, the results changed accordingly, but left much to be desired.
The AI appeared to struggle with recreating finer details like hands, fingers, eyes, and on-screen content.
Nevertheless, if you find an image you like, you can easily use it in a new design within the Canva platform.
What I like:
- Fast
- Easy to use
- Lots of visual styles
- Offers video
What could be better:
- Photo-realistic results
- Depictions of people
- Attention to detail
Best for: Those already investing in Canva Pro. This can be useful for art, vectors, or social media designs.
7. Meta AI
Meta AI is a free intelligent assistant from the parent company of Facebook and Instagram. The company claims the chatbot is “capable of complex reasoning, following instructions, visualizing ideas, and solving nuanced problems,” including generating images.
Pricing: Free for all Facebook users
Test Results
Meta AI is set up as a chatbot and upon entering my test prompt, I was floored.
With each trial, Meta delivered four images — all vibrant, detailed, and in various settings. They also featured diverse groups of people.
Looking closely at the faces, you can see some imperfections. But, for the most part, the images could easily be used in smaller sizes without any concern.
Meta AI also allows you to click into an image to request edits (though this will change the entire image, not just a part, like with DALL-E3).
Or you can animate the images with the click of a button.
What I like:
- Free
- Fast
- Multiple image at once
- High-quality, photo-realistic images (crisp, detailed)
- Accurate depictions of people (mostly)
- Can animate results with the click of a button
- Lots of diversity (both in people featured and in style/setting of image)
What could be better:
- Specific edits
- Images are watermarked
Best for: Small digital graphics for social media, eBooks, or otherwise.
8. Gemini
Gemini, formerly Bard, was launched by Google in December 2023. Similar to competitor ChatGPT, Gemini responds to text prompts as a chatbot.
However, taking a page out of the Google search engine playbook, it can natively understand images, audio, video, and code. In other words, you can upload a video and ask Gemini to summarize it.
Pricing: It's free, but “For Business” accounts are available with Google Workspace. Gemini Advanced is also available with a Google One AI Premium plan for $19.99 per month.
Test Results
In February 2024, the ability to create images of people with Gemini was removed noting “some of the images generated are inaccurate or even offensive.” In fact, the first time I researched this topic, I couldn’t even generate an animated depiction of a person.
But in 2025, the ability returned, and honestly, I was shook by the results.
Gemini processed for about 20 seconds and then delivered a single image in response to my prompt, but honestly, it was the most photo-realistic image I’ve encountered from an AI tool.
The image was crisp and not only were there no issues with facial features or body parts, these elements were extremely detailed. Hair had fly-aways, skin had texture, and hands had veins. It looked like it came straight off a smartphone (in a good way.)
From the result, all you could do was download or share the image, rate it (thumbs up or down), or ask Gemini to regenerate it. There’s no dedicated option to edit, but I imagine you can request the chatbot to make changes as you would in other tools.
In my second trial, the results remained impressive.
I liked seeing the tool introduce a new face and setting, but do wish there had been more diversity in the human depictions. After a few regenerations, I was able to get subjects of different cultural backgrounds, but even after five tries, I couldn’t get a result with a woman.
What I like:
- Easy to use
- Free
- Fairly fast
- Images are extremely high-quality
What could be better:
- More customization options
- Only one result at a time
- Vastly improved diversity, but still room for growth.
Best for: Blog, website, or even print images.
9. Jasper Art
Jasper is a platform trusted by over 100,000 leading brands and agencies including Wayfair, Ulta, and MorningStar.
Its AI image generator is called Jasper Art (only available under Pro plans) and promises users the perfect picture to match their messaging.
It offers high-resolution, 2,000-pixel images, royalty-free commercial use, and unlimited generations, all without a watermark.
Pricing: Starting at $49 per month per seat or ($39 per month per seat with annual payment), but a 7-day free trial is available
Test Results
I learned you can start creating from scratch with “free form” or with a “template” which includes categories like food photography, ink art, news graphic, and storybook photography.
In “free form,” Jasper Art allows you to refine your creation with a mood (i.e., calm, gloomy, or whimsical), medium (i.e., collage or ink), style (i.e., pop art or art deco), keyword (i.e., black & white or close-up), or language.
For this experiment, however, I stuck to “free form” and left all of the refinements empty.
Jasper delivered four images and took just a few seconds, but, to be honest, the results were lackluster.
The “young executives” all appeared older and were men with lighter skin tones. Few women were in the photos, and if there were, they were in the background. This was consistent throughout my trials, so, like DALL-E3, I had concerns about AI bias.
Furthermore, Jasper struggled with recreating features like hands and fingers. One image even appears to have an elf leg coming out of a man’s hip onto a table.
Like other tools, Jasper’s results were photo-realistic, but to confirm, I reran the prompt using the keyword filter “photorealistic.” The results were unchanged.
I also experimented with the styles (specifically pop art and acrylic paint) to see how the tool handled those.
Pop art was true to its name, but Jasper appeared to have difficulty with acrylic paint, delivering images that looked half vector and half photo-realistic.
If you like an image, Jasper Art lets you download it in three different sizes, copy it to your clipboard, or share it to X (formerly Twitter), Facebook, or Reddit.
What I like:
- Easy-to-use
- Multiple images at once
- Offers templates
- Offers high-resolution downloads
What could be better:
- Difficulty recreating human features
- Lack of diversity
- Concerns of bias
Best for: Abstract or creative images. It's worth playing around with if you already pay for Jasper Pro but wouldn’t pay just for Jasper Art.
Pro tip: In “free form, ”Jasper AI offers to “enhance your prompt” with the click of a button. It expanded my prompt and passed the 400-character limit while allowing me to edit before generating my images. While I didn’t use this feature in the experiment, Jasper's assumptions seemed to match the pictures it produced.
Reviewing the more detailed prompts may give you more insight into the image it will create by default.
10. Typeface
Typeface is an enterprise-grade multimodal AI content creation platform for creating on-brand, targeted content for various channels.
It is known for its AI product photography capabilities, but its Visual Inspiration Studio also allows users to pick their inspiration from a library of pre-set images and generate images from text prompts using “Magic Prompt.”
Pricing: Starting at $49/month per use, according to Deepgram. 14 day trial available.
Test Results
To test the tool, I started a “new project” and hit “create from blank.”
From there, I entered my prompt with “Magic Prompt” toggled on and let the tool do its thing.
I left the default settings on, but I could have adjusted the size, generation type (or photo style), as well as details like the audience, things I don’t want to see (via the negative prompt), as well as advanced options like camera angle, lighting, etc.
After about one minute, the tool processed the request and produced a more detailed prompt, followed by four images.
The were detailed and definitely photo-realistic, despite not specifying that. Like a few of the other tools I tested, they looked like they came straight out of a smartphone, right down to the lighting and angles.
Upon deeper examination, there didn’t appear to be any issues with hands or fingers, but the tool did seem to have some trouble with the greater details of faces when I zoomed in.
There also didn’t appear to be much diversity in the style/setting of each image or the human subjects within them.
After running my second trial, the results were similar, but I was happy to see a bit more diversity and a new setting from the first collection.
What I like:
- Truly photorealistic results
- Options to give camera specifics
What could be better:
- Difficulty recreating facial features
- Advanced features only available in the enterprise plan
- Learning curve
Best for: Product photography, professional-quality marketing visuals. Businesses can get some great branded photos and images for ad campaigns, e-commerce store listings, brochures, social media, and more.
11. Getimg.ai
While researching this article, I found Getimg.ai in a Reddit discussion. Redditors raved about it, so I had to see why. With a paid plan, it can generate photorealistic, artistic, or anime-style images, up to 10 at a time.
Pricing: Free for 100 credits per month, or paid plans beginning at $12/month which unlock more credits and features
Test Results
Getimg.ai generates four images by default on a free plan, and it can deliver up to 10 with a premium plan. It’s also transparent about its speed, displaying how long it takes to generate each image. The images I created took under 30 seconds.
My immediate reaction: Zero diversity. Except for one or two subjects, everyone in the photos is a white man and many of them look like they could be cousins, brothers, or even twins.
In other words, Getimg.ai seems to use the same face repeatedly in one image.
If this works for you, the tool lets you like, download, generate similar images, or use them in a design. With a paid plan, you can also convert the image into a video.
As the tool defaults to photorealistic, I once again deviated from my test edit to run the prompt in other built-in styles.
“Anime” delivers some beautiful creative images that are very much in line with what you’d expect from the Japanese style. In contrast, “Artistic” is reminiscent of characters in a video game. I personally expected them to look more like paintings or illustrations.
What I like:
- Generous free plan
- Easy to use
- Fast (and transparent about generation speed)
- Multiple images at once
- Can upload reference image and tell it what not to generate with “negative prompt”
- Multiple aspect ratio options
What could be better:
- Diversity (in demographics as well as the setting of the photo)
Best for: Creative or artistic images. Brands or marketers targeting younger audiences interested in Anime.
12. Fotor
Fotor is a popular photo editor and design platform. Fotor has also launched an AI image generator with a user-friendly interface that is a much easier way for users to generate images. This can greatly help users save time and have fun at the same time.
Pricing: New users will receive four free credits as rewards. Fotor Pro users will receive 100 credits and pay $3.33 for a month.
Test Results
Fotor, like many other tools in my experiment, worked with simple text prompts. All I had to do was enter my prompt, and define my photography style (56 options), dimensions, and the number of images (up to 4), I wanted to generate.
The tool also allowed me to input what I wouldn’t like to see in the form of “negative prompts.”
Upon entering my prompt, Fotor delivered two images — and I was genuinely impressed.
The style called Photography 4 produced crisp, high-quality images.
Did they look like they came out of a camera? No. They definitely appeared digitally generated, but they were detailed, showed no signs of issues with facial or body features, and gave very visually different options to choose from.
For the second trial, I asked the tool to produce 3, 4:3 options.
It followed through with the high-quality results, but the faces did appear a bit more graphic. In both trials, there also did not appear to be much diversity in the human depictions.
What I like:
- User-friendly interface
- Customizable
- Cloud save data access
- Various image styles
What could be better:
- Image details
- Image elements variety
- Diversity of human depictions
Best For: Creative and illustrative images generation.
Put the Best to the Test
What you consider the best AI for creating images is subjective. In my opinion, many of the free tools have more to offer marketers than the paid ones.
I was particularly impressed by Adobe Firefly. It delivers some of the most realistic photos and professional-looking artistic images on the list, and it allows you to edit specific details.
For a marketer who is likely using an AI image generator to create an original image for content or a digital graphic, it more than gets the job done at no cost. On top of its creative prowess, it’s hard to top.
Editor's note: This post was originally published in May 2024 and has been updated for comprehensiveness.
