Seeing is believing. But with generative AI, you have to start with words to get that picture.
Generative artificial intelligence technology has exploded in popularity over the past few years, with a number of creative services emerging that can generate entire images and video clips. This is the “generative” aspect of gen AI. You can generate content from a seemingly simple text prompt. Of course, the actual process is more complex.
AI image generators rely on databases of existing content (text, images, videos, etc.) or gather information from the open web to create new images that satisfy your requests. Anyone with an account and a prompt will be able to create almost any type of image (usually within reason).
There are many AI imaging services available, but not all provide high-quality results or adequately protect your data. They also differ in price and features offered. While AI generators can be fun to explore, they can quickly become overwhelming once you start using them.
That’s where this guide comes in handy. I tested and reviewed multiple AI image generators for CNET, highlighting how new technology impacts digital creators. These services are certainly interesting and serve their purpose, but there is a learning curve. Although it is not sufficient to replace human creators, it does offer a variety of creative possibilities.
Here’s how to create a custom AI image:
Step 1: Choose the best AI service for your project
I have worked with many AI image generators and creative programs. The best program for you will depend on your project, budget, and experience level. Check out detailed reviews and rankings. These are the three programs I recommend.
What is the best AI image generator for you?
CNET’s Editor’s Choice is OpenAI’s Dall-E 3, creator of ChatGPT. Dall-E boasts excellent fast compliance and handles more complex prompts with ease. Conversation flow (meaning you can easily follow up by typing in the chat window) makes editing easier after production. Dall-E is one of the most creative programs we’ve tested. A ChatGPT Plus account ($20/month) is required to access this program. Also, be aware that OpenAI may use your content to train models. However, you can submit a request to stop training your data.
If you’re looking for a free program, we recommend Leonardo AI. Canva recently acquired this, but you can still use it on your website. Leonardo offers more editing tools behind a paywall and a powerful free plan with several artistic styles to choose from. We also offer unique prompt improvement tools that allow you to take simple ideas, rewrite them, and fill in the details to create a beautiful final result.
If you’re on a budget and have no creative experience, check out Canva’s Magic Media generator. It’s very intuitive and easy to operate, making it perfect even for beginners and non-creatives. Canva offers 50 free credits for image generation (these are permanent credits; once you’ve used up your free tier, you’ll need to upgrade to get more credits on the paid Pro plan) and content Automatically opt out of AI training for. .
After choosing a program, create an account and proceed to the prompt window.
Step 2: Create an image prompt
Your best chance of success starts with a well-written prompt. All prompts should include character basics, setting, other elements, and image dimensions. However, to really zero in on the image you want, add a description of the artistic style, aesthetic, mood, and specific colors you want to emulate.
Leonardo may not understand the prompt “cottage-core coastal grandma,” but he does understand “rustic vibes with blues and warm lights.”
Caitlin Chedraoui via Leonardo AI/CNET
AI image generators naturally have limitations in what they can generate. This is to prevent people from abusing these services to create content that is illegal or potentially dangerous or offensive. For example, Dall-E 3 refuses to generate images that request the names of public figures such as President Biden or Taylor Swift. Be sure to review the service’s safety policy before generating.
Prompt engineering (the process of creating appropriate prompts to achieve desired results) gets better with practice. It also varies with each image generator. For example, Canva tends to be better with simple prompts, while Dall-E recommends longer, more detailed prompts. The best way to understand a program is to try it out. For more tips and tricks, check out my complete guide to image prompt engineering. Once you have created the best prompt, proceed to generation.
Step 3: Edit the image as desired
Even with the best generator, you may need to edit your images after they’re created. Depending on the service you use, you may have access to editing tools or the ability to request changes. Be careful not to over-edit your images. I found that the service could only handle so many changes before the image became unstable and ultimately unusable.
Take this image I created on Midjourney as a warning of what can happen if you over-edit. Sometimes it’s best to start fresh.
Caitlin Chedraoui, via Midjourney AI/CNET
Post-generation editing tools are one area where AI services tend to shine or fail. If your service doesn’t offer them (or doesn’t work well), try editing prompts instead of images. In my experience, you have a better chance of success doing so than incrementally editing an existing image.
Step 4: Credit images properly
Labeling images as AI-generated content is very important. As AI image services improve, it is essential for all creators to clearly identify the origin of their images, especially distinguishing them from human-created art and photography.
In the bottom left corner, Meta AI will watermark all AI-generated images.
Caitlin Chedraoui, via Meta AI/CNET
Some services automatically add watermarks or other markings to indicate they were generated by AI, but many do not, including Dall-E, Adobe Firefly, Midjourney, Canva, and Leonardo. When sharing images online, be sure to include the AI service you used for captions, credits, and alt text. If you share on social media, you can also add a disclaimer to your post. Instagram and Facebook may flag your meta as containing AI content, but adding a line to your caption can help avoid confusion.
Frequently asked questions and other considerations
At the time of publication, the use of AI image generators is legal. However, please note that there are important legal and ethical considerations when creating AI images. Here are some frequently asked questions.
AI image services like Midjourney and Dall-E leverage existing content to create images. This content is typically created by humans, both in databases and on the open web. As such, some creators and artists have major concerns about how these programs are designed and trained. This has led to an uproar in the industry and several high-profile lawsuits, including an ongoing lawsuit against Stable Diffusion, an AI model used in a popular generator.
If you’re concerned about using a service that may unfairly treat creators, choose one that clearly discloses how its models are trained. Some AI services rely on public databases, such as Adobe Stock or Getty Images, where creators contract with companies to input their work into these collections for specific uses. Adobe Firefly, for example, says it only uses content from the Adobe Stock public database and images with expired copyrights. You also can’t train based on content.
As these AI services evolve, the legalities and ethics associated with them are likely to change as well. If you want to become a regular AI art creator, make sure to stay up to date with the latest requirements. This is explained in the AI Atlas guide.
