Prompting makes perfect: Award-winning AI artist Nóra Bereczki on new frontiers for photography

Is an AI-generated image really art? What role will AI play in schools and will machines take our jobs? Award-winning Digital Umami creator Nóra Bereczki gives her artistic impression on the creative processes, photography and education. Artificial intelligence is transforming our entire world, from street billboards to brain surgery. Art is no exception. An ever-growing number of creators use AI models, while a lively debate is under way over whether works created in this way can even be regarded as artistic creations. Enter Nóra Bereczki, an AI artist working under the alias Digital Umami who's attracting increasing attention. Her video series that bring Hungarian poets and writers to life has appeared widely on social media, and this summer she won an image-generation software contest from among 52,000 entrants, taking home the $100,000 grand prize. Her winning series is titled Dolce Impossible, inspired by the Mediterranean summer. It consists of ten images that depict typical holiday scenes, which many people can relate to, but each of them contains a surreal element. For example: What is the difference between an average AI-generated image and art? Why is it dangerous to see artificial intelligence not as a tool but as a partner? What role can AI play in education? How should we prepare our children for the new world shaped by artificial intelligence? And will machines really take our jobs? These are just some of questions we discussed with Nóra Bereczki. First of all, how did you first come into contact with artificial intelligence? Professionally, I work as a marketing consultant, and I am very interested in technology. My last workplace before I went on maternity leave was a robotics and industrial automation company. My husband also works in AI, and we talk a lot about the topic, follow what is happening and how the technology is developing. AI tools broke into the wider public consciousness in 2022, and that was when I began to think about how they could be effectively built into a company's marketing strategy. Alongside marketing, I have been involved in analogue and digital photography for more than 20 years, and image-making is my passion, so I started testing these AI tools separately as well. Hobby projects followed, I kept a close eye on how the technology was evolving, and the hobby gradually grew into something bigger. If I understand correctly, you use AI to realise your own ideas for images. Have you finally found a tool that lets you do anything you can imagine? You're on the right track. When I walk down the street, I constantly see frames and think in terms of compositions. There are a huge number of such ideas in my head, and many of them I cannot realise through classical photography. AI, however, is perfect for expressing these ideas visually. A good example is one of the images in my award-winning series. On a beach, a huge, human-sized lobster stands facing a three-year-old girl. The scene itself had already existed in my mind. Of course, it would be possible to try to create it as a traditional photo production, but it would require substantial organisation, post-production and budget. With the help of AI I was able to realise the visual idea I had imagined. Perhaps this is also why I don't see AI as something that replaces photography for me. It has become another tool on the palette. I like mixing different technologies and genres; for example, I often combine my own analogue photographs with AI. I always try to decide which tool suits a particular idea best. Sometimes that tool is AI, and sometimes it is not needed at all. Has it ever happened that you had an idea, but AI overrode it and you liked the new version so much that you decided to go with that instead? It's one thing when someone is just experimenting with image-generating tools, types in something like "blue dog", the model produces something artistic, and the user likes the result and feels a sense of achievement. And it's another thing when someone sits down in front of AI with a specific vision, and the question is whether they can carry that vision through. In my view, an AI-generated image becomes a human creation when the artist is able to carry through the vision they imagined. When they do not leave the outcome to the random results produced by the model, but really get what they originally wanted. I also have an AI image on my bookshelf, and it really was created in the rather random way you described. In your case, I imagine producing such an image is quite a long process. If someone makes an AI-generated, random image that brings them joy, that's great. But I generally approach things with a fairly clear picture in my mind of what I want to see. I know what elements I want the image to be built from, what the composition, lighting, texture, aspect ratio or even the camera angle should be, and I try to pass all this on to the model as precisely as possible. From there, it becomes an iterative process: I look at what it understood from what I imagined, and if something isn't right, I refine it. Sometimes the desired result appears very quickly, other times it takes many rounds. For me the crucial thing is not to let the first interesting generated image determine the direction I go in, but to have a vision and consciously steer the outcome towards it. Does that require very long instructions? Not necessarily. A prompt doesn't become good simply by being long. It matters more that it is well thought-through and clear. But to really get back what I imagined, you have to decide on a lot of small details. For example, it can happen that I specify almost everything except that the scene should take place in daylight, and the model generates a night-time image. In that case, you need to fine-tune the instructions and generate again. At the same time, you have to make sure not to overload the prompt, because beyond a certain point too many instructions can actually worsen the result. And of course different models vary greatly in how precisely they interpret textual instructions. When social media became widespread, the mantra was that everyone could now be a content creator. Thanks to AI this is now exponentially more true. But is there a threshold someone has to cross in order to create works of art with artificial intelligence? For instance your background in photography? In technical terms, I think the entry threshold is now very low. AI provides a sense of success incredibly quickly: you type in a few sentences and within seconds you get a striking image. As a result, vast amounts of content are being produced, and it's no coincidence that the term "AI slop" has appeared to describe low-quality content generated at scale. The artistic threshold is, in my view, a completely different question. What I see is that behind the works that really manage to stand out from this noise, there is very often some kind of relevant background knowledge. It might be photography, film, graphic design, painting or another visual field. The knowledge, experience and visual sensitivity someone has built up over the years matter a great deal: how composition, light, colour and rhythm work, or how to tell a story visually. For me, my photographic background is clearly a huge asset. At the same time, one of the most exciting things about AI is precisely that it can put tools into the hands of people who have always had a creative mindset or sensitivity but never found the right medium for it. I know someone who used to work in a completely different field, yet always felt a strong need for self-expression. In AI they found the tool that finally lets them visualise their ideas, and they are producing fantastic work. Hungarian creators, incidentally, are very strong in this field internationally as well. Dávid Szauder, for example, attracts significant international attention, and then there is Laci Gaál, who lives and works in Vietnam and is also internationally recognised. It is telling that they too have a prior visual background: Dávid Szauder comes from fine art, while Laci Gaál comes from film. One of your projects, the AI videos presenting Hungarian poets and writers, has already been used by several teachers in literature classes. The obvious question is: in your view, how can artificial intelligence be used in education? I would divide this into two parts. One is when the teacher uses AI. I have spoken to several teachers who are specifically interested in the technology and are creating very exciting things with it: quiz apps, interactive teaching materials, or tools that help visualise scientific phenomena that are hard to imagine. To do this, however, the teacher has to understand the tools and their limitations. If we want to embed the use of AI in education at system level, I think we first need to prepare teachers. Not only in how to use a model, but also in when it is appropriate to rely on it, how to check what it says, and what data protection, copyright and ethical issues arise. This is a particularly difficult task with a technology that develops faster than we can adapt to it. At the same time, it is necessary to prepare education systems for AI's presence, because pupils will encounter it in their everyday lives and later in the labour market. Across the world, in several countries and education systems, initiatives have already been launched on how to incorporate AI and AI literacy into education. And what is the other case? The other is when the child themselves uses AI for learning at home, and here parents have a crucial role. As a parent, this concerns me a great deal: at what point would I first put such a tool into my own children's hands? In part this is a question of age, but it also depends on what skills need to be in place beforehand. I think a child first has to learn to study without AI. They need to experience what it is like to delve into something, think through a problem independently, search for information, make mistakes, try again and think critically. And of course they also need to learn what AI is, how it works, in what ways it can be useful, and what its limits and risks are. That is why, for example, in the early years of primary school I don't yet see a place for simply sitting children down in front of ChatGPT or other generative AI tools. Once those basic competences are in place, I think AI can offer huge opportunities, especially in personalised learning. It can explain the same material in different ways, adapt to the child's pace, offer different examples to those who didn't understand something the first time, or suggest alternative learning approaches to pupils for whom traditional methods are less effective. For a child with ADHD, for example, it can help break the material down into shorter units, make it more interactive or link it to their interests. So in my view the question is not whether we should use AI in education, but when and for what purposes, and on what foundations. Let me quote you from an earlier interview: "The question is not whether the tool is artificial, but whether the thought behind it is human." Are you saying that anything that is not born of a human thought but of artificial intelligence cannot be an artistic work? That it cannot be art if I tell AI to create a beautiful work that expresses the pain of passing time? It's hard to say, because what you described could easily slip into the category of experimental art. Many works of art have been based on a random situation like that. But even in such cases there is, once again, a human thought: the artist's aim was to have the artwork emerge with the help of chance. In my own creative process, the starting point is usually some inner tension, a feeling, thought or question that I want to release in some way. Then comes the vision of how I want to show it, and I choose the tool to do so, which may well be AI. So the idea comes from the human, and AI is the tool for realisation. It's like using a robotic arm to lift something heavy. Many people, however, don't treat artificial intelligence as a tool but as a companion. Yes, and for me this is where one of the truly problematic aspects of the technology begins. I have done interviews where people referred to AI as a creative partner, and I have to admit that gives me chills. I very consciously see it as a tool. Creative intent, decision-making, control and ultimately responsibility must remain with the human. We can already see people giving names to the models they use, talking about them as personalities, or feeling emotionally shaken when an update changes or removes the version they were used to. In other words, a real emotional bond can form with something that is, in fact, a technological system. I consider this dangerous because AI can imitate human communication in a highly convincing way without there being a person on the other side who feels, understands or cares about us. The better it becomes at this, the more important it will be for us not to lose sight of that distinction. There is, for example, a deeply disturbing video in which a grandmother receives a ChatGPT subscription, and by the time her grandchild explains that she isn't talking to a real person, she already wants to invite them round because they've become such good friends. These are things I find very worrying. And we can't really blame the grandmother for believing she was talking to a human being. The technology is developing much faster than we are learning to live with it as a society. We're long past the point where you can spot an AI-generated image because a person has six fingers. It is becoming ever harder to decide, on the basis of our senses alone, whether something is real. It is hugely important that we prepare our children for this new world. I already talk to my kids about the fact that just because they see something in a picture or a video doesn't mean it really happened. In their world, the basic assumption won't be "I'll believe it when I see it". Perhaps the most common anxiety around artificial intelligence is that it will take away our jobs. How real do you think that threat is? I think it's an absolutely valid concern. We are facing very serious changes in the labour market; it's just that no one yet knows exactly how extensive they will be. Stanford University recently published a study showing that in the United States, job prospects for young people in professions most exposed to AI have already clearly worsened. The researchers looked at what percentage of 22 to 25-year-olds qualified for a given profession are employed, and found that in the summer of 2025 there was a 15 percent gap compared with professions less affected by AI, and by the summer of 2026 that disadvantage had already grown to 19 percent. At the same time, of course, new positions that did not previously exist are also emerging; I see more and more job ads specifically related to AI. And how should we respond to that challenge? What should we do, if for instance as journalists we are worried about whether we will have jobs in the long term? What I think we absolutely must not do in this situation is stick our heads in the sand and tell ourselves things will somehow work out. It is worth regularly thinking through how our own profession might change, what new knowledge we will need, whether we should retrain or even look for a new professional direction in the longer term. And I think we should get to know these technologies even if we disagree with them or are unhappy about their spread. They will affect our work and our environment regardless. As parents, we also have to keep in mind how we prepare our children for a world and a labour market that will probably be very different from the one we grew up in. We need to keep an eye on the trends. No one knows exactly what will happen, but it is already clear that we are headed for a great many changes.