In recent years, tools to produce AI-generated images have become widely available – and have exploded in popularity. Like so many industries, graphic designers are now grappling with the impact of artificial intelligence on their work and livelihoods. So, what’s the state of play, and what does it mean for the future?
What are ai-generated images?
According to the Visual Storytelling Institute:
In a nutshell, it’s an image created by artificial intelligence (AI) algorithms … The AI learns the visual structure, style, and text labels for a lot of images. Then, upon receiving a text prompt can generate a new image. The logic is simple. If a machine can recognise an object such as a cat, it should ideally also be able to recreate a cat.
Today, there are a range of machine learning-based image generators that convert text prompts into visuals – including Midjourney, Stable Diffusion, and DALL-E 2.
These tools make it possible to create unique images and graphics from a simple description or other forms of input. Here are some examples of text prompts – and the results generated by AI programs in response.
Graphic designers and ai:
An evolving relationship
So, how does the proliferation of AI-powered image generation affect modern graphic designers? As brand identity, graphic design and marketing professionals, the Mariart team has summarised the opportunities and challenges of AI-generated images for our profession and our clients. And one thing is certain: it’s a dynamic and fast-changing landscape.
AI-generated images:
The opportunities
Need for speed
With the right text descriptions, an AI image generator can create visuals in seconds. This speed has obvious benefits for graphic designers, especially for image-heavy projects and tight deadlines. AI-generated images can also be used to create mood boards and mock-ups easily and cheaply.
Creative fuel
AI-generated images can be a great source of fresh inspiration for graphic designers, and a fast and fun way to explore new approaches, styles and themes. The ability to generate a wide range of unique visuals quickly can be the key to overcoming a creative block or difficult challenge while accelerating the brainstorming phase.
By inputting different parameters, AI tools can generate new ideas, with the freedom to experiment, refine, and streamline the iterative creative process. Instead of developing everything from scratch, designers can use AI to explore different typography and colour schemes or generate templates they can personalise.
Flexibility and ease
AI image tools provide more ease and flexibility in graphic design. They’re great for creating designs, visual elements and layouts for various purposes. Depending on quality, they can be used in many different formats, such as website banners, flyers and even printed documents.
For most graphic designers in this competitive industry, it’s now much simpler to add or remove objects (even usually difficult features, like clothing) and replace or extend backgrounds – all while taking into account the lighting of a scene and without a painstaking despatching process — something many graphic designers appreciate.
But this is just the beginning, as these tools can effectively communicate unusual or unique scenarios without having to endlessly comb through stock images or organise an expensive and elaborate photoshoot. All this can save time, boost efficiency in the design industry, and add capacity to take on more work for more clients.
Customisation
Graphic designers can tailor AI-generated images to match specific project requirements, using prompts that include style, colour, size, and aspect ratio to create logos or artwork that suit their needs. These tools can provide unique solutions for unique client challenges, given they create entirely new images that do not exist in reality.
Accessibility
AI-driven tools, combined with accessibility-focused design principles, can improve the inclusivity of design. AI can automatically generate descriptive alt text for images, making websites and digital content more accessible to visually impaired people. They can be used to adjust aspects (like colours or contrast) so images are more distinguishable and can generate natural language descriptions.
AI-generated images:
The challenges
Tricky text
AI image generators struggle to comprehend text symbols (such as letters, numbers and characters). Typography will often be represented as unintelligible symbols, so if an image has (for example) a clock in it, designers still need to be able to composite the image with non-AI typography within Photoshop.
Body issues
AI doesn’t understand the anatomical reality of being human – and we often find that human bodies in AI-generated images have strange and inaccurate proportions and limb placements. The software can struggle to represent the subtleties of ageing, feature unrealistic hair and skin textures, and show people in unnatural or awkward poses.
Not handy
AI-generated images are notorious for ‘bizarre’ hands: with extra digits, fingers sticking out of palms, and two or more hands fused at the wrists. The reason? The software finds it hard to ‘learn’ what hands really look like and generate them realistically. Hands are intricate – and pictures in training data sets are often more focused on faces, while hands are often folded, gesturing or holding something. Fewer recognisable patterns makes it harder for AI to recreate them correctly, although more advanced software is starting to improve the situation.
Resolution and quality
With the right prompts, most tools are capable of high-quality AI-generated images. Their size can vary, with some limited to around 1024 pixels (or around 1 megapixel), requiring the use of upscaling tools for use in larger print applications.
Power of the prompt
It takes skill (and trial and error) for graphic designers to master the best prompts to get the best visuals. In a recent article, Hootsuite points out an obvious but crucial point: AI generator tools are robots and don’t understand the world as we do.
“AI art generators don’t know what an owl looks like in the wild. They don’t know what a sunset looks like in a physical sense. They can only understand details about features, patterns, and relationships within the datasets they’ve been trained on. Prompting for a “beautiful face” is not very helpful. It is more effective to prompt for specific features such as symmetry, big lips, and green eyes.”
The best prompts will generate the best results, faster. Think clear and detailed (but not confusingly long), with attention to composition, colours, lighting, perspective and artistic styles – as well as background and settings. According to an article by Dr Marcel Scharth from the University of Sydney Business School, AI-prompting skills will become more in demand and “prompt engineering is essential for unlocking generative AI’s capabilities.”
Human Touch
AI-generated images and logos are created by algorithms without the depth of emotional intelligence, judgement, brand insight and creativity that human photographers, artists and designers provide. On their own, AI tools can’t truly understand or reflect a client’s unique brand personality – and business goals and objectives.
it comes to photo-realistic images of humans, there are some clear limitations. AI-generated faces can be “too perfect” with unnaturally symmetrical faces.
AI-generated images can sometimes appear flat and lack depth.
The details
When it comes to AI-generated images, graphic designers need serious attention to detail. All outputs require thorough checking, evaluation and editing. Problem-solving in this realm is essential, as is the ability to receive feedback and apply styles as needed. In many cases, prompts need to be fine-tuned many times to generate the right results – and images will often need further work and manipulation.
Legal & ethical considerations
for creative professionals
generative AI systems being trained on data scraped from the web, like images, is murky ground for copyright.
According to the Professional Association for Australia’s technology sector
There have already been a number of AI-related lawsuits (still pending) in the US and UK. To date, it appears that no lawsuits have been filed in Australia alleging copyright infringement against users of generative AI systems. However, the primary concerns regarding copyright infringement involve individuals or organisations that have trained their generative AI systems using copyrighted material.
Users who are simply utilising AI tools to generate images don’t appear to be at risk of any infringements as long as the tools they use aren’t built on copyrighted material. On the other hand, an Australian Copyright Council submission suggests that “using generative AI may pose copyright infringement risks in relation to the output of AI tools”.
For now, designers and amateurs alike are able to utilise and monetise works created with these tools with little risk, but there is no doubt the issue will be thoroughly reviewed in the future as the use of AI-generated content explodes even further.
When it comes to ownership of the images actually generated by AI, Australia does not specifically provide protection for ‘computer-generated works’. Under current Australian law, a work can only be protected by copyright if a human author contributed “independent intellectual effort”.
This situation is somewhat like using a stock image for various purposes. Popular stock images are often used by multiple businesses and graphic designers, blurring the lines of ownership. Each business or graphic designer using these images has paid to do so, often with the rights to use them for monetary gain in social media posts, websites, or other visual communication tools. This means that works solely created by AI may not be protected by copyright, leaving doubt around the ownership of these images.
It’s an evolving situation in the design industry, so as professionals wanting to innovate and automate our artistic practice and communication skills, we must be committed to enjoying the benefits of AI while ensuring an up-to-date understanding and compliance with all existing copyright and IP laws.
Of course, there are other issues that require caution, and it is important for any good graphic designer to be aware of them. Many clients are fearful of AI-generated content with ethical concerns around authenticity, deepfake technology, and writing that perpetuates social stereotypes. There are also issues around AI models and bias around age, race, and gender.
A study by Bloomberg included a thorough analysis of more than 5,000 examples of images created by a prominent AI image generator – and found that it takes racial and gender disparities to extremes. The results are worse than those found in the real world. It is hoped that in the same way we tackle other barriers, bias will reduce as AI models advance in this creative world, but for now, the challenge continues.
Graphic designers and AI:
Other tools
In addition to AI-generated images, there are a raft of AI tools to help modern graphic designers, including large language models that can generate written content, like ChatGPT.
While this can be helpful to generate initial ideas and structures for copywriting – as well as filling “gaps” in design – AI-generated text is often verbose, repetitive, impersonal and outdated. It needs careful review, finetuning, editing and quality control.
Other useful AI tools can assist in layout and composition, augmented and virtual reality experiences, feedback and testing, animation and video production, data visualisation and collaborative work.
AI-generated images:
Real-work examples
Here at Mariart, we know how to use AI image generation to solve problems – with great results.
MediHerb, an Integria Healthcare brand, provides healthy recipes as part of their free resources. The challenge? The difficulty of sourcing stock images with the specific (and sometimes uncommon) combination of ingredients – like carrot, beetroot and coriander. The solution: quick, easy and colourful AI-generated images that perfectly fit the bill!
On rare occasions, MediHerb also needs to represent health treatments in specific settings, like acupuncture being performed in a non-clinical environment.
With nothing suitable in stock libraries, we combined AI tools with the right prompts and design work to generate an on-brand image that’s ideal for purpose.
Graphic designers and AI:
Collaborative approach
AI-generated images and the tools that create them provide both opportunities and challenges for graphic designers and brand agencies.
They can streamline workflows, inspire and test ideas, add flexibility and accessibility – and so much more. Of course, like all AI, there are also downsides, concerns and evolving considerations to work through.
Here at Mariart, we embrace the opportunities of AI-generated images while firmly understanding there is no replacement for human creativity, intelligence and artistry.
It’s fantastic to have AI in our toolkit: helping boost productivity, sparking creativity and new ideas in the early project stages, and generating large numbers of visuals when needed.
However, AI will never substitute our expertise, strategic approach, client relationships and deep brand insights.
In short, AI complements and enhances our capabilities for clients – and we agree with Ginni Rometty that perhaps a better term is “augmented intelligence”.
Some people call this artificial intelligence, but the reality is this technology will enhance us. So instead of artificial intelligence, I think we’ll augment our intelligence.”
Ginni Rometty, Former CEO of IBM
Want to know more – or work with us to create powerful visuals, communication and campaigns to build your brand and truly connect? Simply get in touch for your first, free consultation.