Neural Network that generates Photorealistic Avatars
Hi designers! I have been experimenting with various Neural Networks for over a year now. I can say that they have already become a part of my everyday life. I actively use Neural Networks for generating text and graphic ideas, searching for references and finding inspiration. Of course, at the moment, we cannot say that Neural Networks are highly intelligent. At this point, they are more like children who are still learning, exploring the world, and are in the process of forming.
Many experts are already sounding the alarm and saying that Neural Networks pose some danger, as they replace manual labor, and many jobs could be at risk. Technical professionals such as designers, developers, marketers and copywriters may be out of work in the near future.
Actually, I can’t share this point of view. If we talk about designers — in my opinion, there will always be work for in-demand specialists. On the contrary — our activity will change somewhat and Neural Networks will allow us to avoid monotonous routine work, which used to take a lot of time. The fulfillment of tasks will become more creative and dynamic.
If we talk about the UX field — Artificial Intelligence will definitely not be able to work out all the Interface Interaction Scenarios, Users Journey Maps or conduct UX research any time soon. But Neural Networks will help us avoid a large number of small routine tasks, such as selecting colors and fonts, searching for references, working with composition, grid and so on.
Of course, there is a risk that some specialists will be unemployed. This will probably apply to low-skilled specialists. It is very likely that Neural Networks will be able to completely replace the work of Designers of banners, posters or covers for Marketplaces. But such work was partially automated before the advent of neural networks, for example by stock services such as Freepik or Figma Community Templates, and also with the advent of simple online graphic editors such as Canva.
So, I think that designers of simple graphic elements, who do not think about user interaction, develop marketing strategies or create technically complex layouts will sooner or later find themselves out of demand. But for highly qualified specialists (for example, designers of user interactions, communication designers or designers of complex print publications) Neural Networks will only help in their work.
Today I would like to take a look at another Neural Network that recently came to my attention and seemed to be quite an interesting option for some creative tasks. The service is called Artflow.ai and allows you to generate photorealistic avatars based on uploaded photos.
I must say that I didn’t bother much and didn’t take photos for this service specially. I just uploaded the latest photos from my iPhone. If you take a photo in accordance with the rules of the service, on a monochrome background — probably the result will be even better. But in general, what I got already surprised me.
The photo processing itself takes 2–3 hours. I just left the browser tab open and went about my tasks. After a couple of hours everything was ready and the system offered me two different characters. The first character is a wizard (magician, spellcaster) and the second character is a blogger (or some technical specialist). The result just amazed me! My avatar in the character of a wizard, in my opinion, turned out very similar to my real photos.
The blogger character, I think, came out less successful. Overall, the face seems to be similar, but the head and body proportions look a bit odd. Write in the comments — what do you think about it? Do we look alike?
Next, I started experimenting and generated a couple more variants. The system allows you to enter a text prompt — describe emotions, clothes, background environment. And, based on this prompt, it generates 4 variants of the photo. In the first request I described a portrait of a serious person, in the second — a portrait of a smiling character on a light gradient background.
In my opinion, the result is worse than in the test images. When the Neural Network tries to overlay different emotions on your avatar, the result is not quite natural. For example, a smile looks a bit plasticine and unrealistic. And in the case of a serious person — the system made me older, and my facial features and body proportions rougher and more masculine.
Overall the result is interesting. The service works fast enough, as far as possible, and makes the photos more or less close to reality. I don’t have a feeling that this is a plastic face generated by a Neural Network. It’s more like a photo retouch. When you register, the system gives you about 100 credits, for which you can generate quite a few different photos. When the credits run out, apparently you have to pay for a subscription. But this is more than enough to experiment and understand whether you need such a tool or not.
Where do I see this service can be used? It will definitely suit photo- and video-creators, bloggers and experts who don’t want to make a photo shoot every time in order to get a cover for the next video or social media post. If you upload high-quality source photos and choose the right prompts, the Neural Network will surely produce very high quality results quickly. Perhaps your audience will not even be able to distinguish the generated photos from your real ones. The function of creating a video, where the avatar can speak with your voice and move its lips, still looks rather weird and not suitable for everyday use so far. But these technologies will probably be improved as well. What do you think about it? Let me know your thoughts!
A Message from AI Mind
Thanks for being a part of our community! Before you go:
- 👏 Clap for the story and follow the author 👉
- 📰 View more content in the AI Mind Publication
- 🧠 Improve your AI prompts effortlessly and FREE
- 🧰 Discover Intuitive AI Tools
