WeSmile creates a unique event experience (with AI)
The photobooth, an indispensable element at your festival or event, to create a unique memory. Can't that actually be made just a little bit more fun? Yes indeed, thanks to generative artificial intelligence (AI), a lot is possible. At least that is what Fontys ICT students Juliusz Dokrzewski and Yung Hui Yap thought, when they started WeSmile; a unique AI-driven photobooth that generates unique photos.
Photo booth
Yung Hui Yap came up with the idea during an internship at Group M. Indeed, there she regularly visited company parties, where there was always one regular guest: "Every time there was a photobooth at the event and everyone was very enthusiastic about it. Meanwhile, within my internship, I saw that personalisation and generative AI offered a lot of opportunities and I got inspired by that. Couldn't the photobooth be more fun with that feature?"
Personalisation
The big challenge in marketing is addressing your target audience in a way that suits them. That is personalisation, explains Yung Hui Yap: "Suppose you are a big car brand, and you want to advertise your new car in London. Then you want a poster with London in the background, because that way people recognise themselves better in the product. Generative AI, like Dall-E, makes that a lot easier for companies. We brought that idea to the photobooth." For that, UX designer Yap formed a team with developer Juliusz Dokrzewski. In the nick of time. The result has already been tested at Night of the Nerds, an annual event for young people with an interest in technology, and Fontys ICT showcase event Innovations Insight. "How it works is that you take a picture and then have it generated based on specific questions, so your result can be linked to the event. We are doing a futuristic event soon, so we link images of robots, Atlantis or Cloud City to the treatment of the image," explains Yap.
Deadline
Juliusz jokes that he is the brains of the project, while Yap came up with the idea: "She spoke to me about 3 days before the deadline to submit your project. We decided to go for this and it worked out just at the last minute. Our education fortunately gives us the freedom to work on our own project." Juliusz and Yap follow Open Learning. This is a form of learning in which students can complete their studies, but also take one or a few semesters in a minor or specialisation. In this, they determine their own learning objectives and projects, and are guided in this by lecturers on a weekly basis. Juliusz: "That's why I chose this. Your coach helps you, of course, and in this case we had Group M as a stakeholder. So we had freedom to do what we want, but also all the guidance we needed."
Challenge of Generative AI
As a developer, Juliusz's job is to get the photobooth working too, and that was quite a process: "We used Dall-e as a model, but we had quite a few iterations with that. I am now very good at what is now called 'prompt engineering' and added different technologies to make our application better. In addition, the physical setup and interface was also important and we had to develop that too." So that's where the biggest challenges are, he explains. The hardware is heavy and not practical for transportation, and using an open source model brings challenges: "We wanted to use a Raspberry Pi for our photobooth, but it lacks the computing power to generate the images. That's a concern. We are now dependent on technology like Control Net and Loras for the result, and there is still a lot of tweaking to do in that. Actually, such a generative AI is like a synthesiser; there are lots of knobs to turn."
WeSmile is now a startup in the Fontys ICT InnovationLab. Stakeholder Group M is the first customer. In the coming semester, Yap and Juliusz will further develop their product within Open Learning and explore how to scale up. Coming soon to an event near you? More info here!