You’ve probably seen AI-generated images circulating on social media. These pictures range from artistic selfies and realistic mimics of famous artists to comically warped photos of humans. Text-to-image generators are already used by millions of digital creators and artists. However, these systems are not perfect and can produce images that are flawed or unrealistic.
AI image generating technology has come a long way, but reviews suggest there is still work to be done before this technology is perfected. What are the key AI image generators you need to know about? How could you incorporate AI image generators into your work? Most importantly, should you bother?
What is an AI image generator?
An AI image generator is a type of algorithm that automatically creates new and unique images. Known as a generative model, the algorithm is trained on a vast dataset of images and uses this data to produce new images.
How do AI image generators work?
These tools are trained using large sets of data. Using deep learning, the algorithm learns how to recognise different types of objects, colours and scenes. Different models use different systems, such as generative adversarial networks (GANs) and variational autoencoders (VAEs).
In order to use an AI image generator, you can input text or images. To create new images based on this input. It may produce a realistic image, or something bizarre and unrealistic. The output depends on what data was used to train the AI and what information is given as a prompt.
What are some key examples of AI image generators?
There are many AI image generators available online, each with its own set of features and capabilities. Three of the most popular examples are Midjourney, Dall.E and Stable Diffusion.
We asked each of these generators to produce images based on the same prompts. As you can see, they produce a variety of responses, each with their own merits and drawbacks.
First we focused the prompt on scenes and colours, asking each generator to produce an image of ‘a man with a red and orange jumper on the beach holding an orange’.
Midjourney produced a range of images that fit the brief well – these are artistic generations and all look proportionally accurate.
Dall.E produced images that fit the brief in terms of colour and scene, but the man in each image is somewhat warped and unrealistic.
The Stable Diffusion generator produced images that are very warped, and failed to incorporate the orange from the prompt into the images.
Next, we gave a more complicated prompt, using location-specific information and requesting the AI to generate language within the image. We used the prompt ‘woman in South Africa standing next to a sign that says “Welcome to Cape Town”. This produced interesting results.
Each generator seemed to understand the context of South Africa broadly. Yet, the people in each image are warped and unrealistic. Most interestingly, the AI was unable to generate language on the sign. Each generator produced a sign with random letters, some close to the suggested phrase but not quite there.
This experiment showed that whilst AI can produce some useful and realistic images, the results are often warped and unreliable.
How can you incorporate AI image generators into your work?
You can use AI image generators to create visual content for your website or social media. Use these tools to create eye-catching infographics, illustrations and marketing materials. This is useful if you’re not an artistic or creative person – the AI image generator can create something for you.
If you’re a creator struggling to come up with ideas, AI tools could help by creating new concepts for you to develop. For example, a video game designer could produce a range of images from a basic prompt and then develop these into plans.
One of the benefits of AI image generators is their ability to quickly create a large number of pictures with minimal human input. This can save time and resources for individuals and businesses. You can also use prompts to adjust images as you go, rather than having to start manual production all over again. This is particularly useful if you are in the development stage of a project and need a range of image samples.
Should you use AI image generating tools?
These tools are useful but they are not an easy trade for human creativity. As shown above, they can sometimes produce results that are not what the user intended. Concerns have also been raised about misinformation, consent, biases and copyright protections.
Prompts can be interpreted differently depending on the available data. It’s important to carefully evaluate the output of AI image generators. You may need to make adjustments and try different tools to make sure the images are suitable for your needs.
Connect with the world and read about the latest news and current affairs. We share ways to stay abreast of the latest science and technology, as well as breaking news stories that you may need to know about.
Follow us on Facebook, Instagram and Pinterest for more articles, videos and content to keep you inspired.
Mobimeme offers content marketing, SEO, analytics, social media management and expert direction in the digital sphere. Building and growing online audiences for your business is what we do best. Get in touch with us to find out more about our package offerings and how you can improve your website and following.