ArtFusion Interactive AI Portrait Kiosk
A portrait experience made for exhibition visitors, not for a notebook.
Built the technology for a kiosk rented by exhibition clients. A visitor could take a photo, receive a personalized portrait with the client's visual identity, get a print on site, and open a digital copy through a QR link.
The system had to work in public, with changing lighting, groups of people, a touchscreen, a printer, and no developer standing beside it to repair every failure.
The visitor flow
The PyQt6 application guided a visitor from the camera to the form, generation, printed result, and QR delivery. Touch input was available, but hand gestures also allowed someone standing away from the screen to start and control the experience.
The generation work
The image pipeline used Stable Diffusion with ControlNet and LoRA so the person could remain recognizable while the visual treatment changed for each exhibition. I tuned the pipeline for low light, several people in one photo, and generation in under 30 seconds.
The product around the model
The backend handled sessions, telemetry, forms, collections, user data, content delivery, and QR access. I also specified the Linux workstation, GPU, cameras, touchscreen, and thermal printer, then handled deployment and on site reliability.
selected tools
- Python
- PyTorch
- Stable Diffusion
- ControlNet
- LoRA
- PyQt6
- FastAPI
- OpenCV
- Docker