AI Art Installation Combines SDXL-Turbo and YOLOv8 for Real-Time Interactive Creativity

In a groundbreaking fusion of computer vision and generative AI, a new open-source art installation is redefining how audiences interact with artificial intelligence in public spaces. Developed by independent developer r_giskard-reventlov, the project integrates Stable Diffusion XL Turbo (SDXL-Turbo) with YOLOv8 object detection to create dynamic, real-time visual experiences that respond to user input—no GPU required in its most accessible form.

The system, detailed in two GitHub repositories—selfusion-pi and sdxl-turbo-api—allows participants to alter the AI-generated imagery on the fly via a web API. This means that in gallery settings, museums, or even public squares, groups of people can collaboratively shape the output by typing prompts that instantly manifest as high-quality visuals. The installation’s brilliance lies in its dual architecture: one version runs on a $35 Raspberry Pi, democratizing access to cutting-edge AI art, while the other leverages a GPU-powered server for higher throughput and resolution.

At its core, YOLOv8 acts as the sensory organ of the system. It analyzes live camera feeds—potentially capturing audience movement, gestures, or even facial expressions—and translates those inputs into semantic prompts for SDXL-Turbo. For example, if a child waves their arms, the system might interpret the motion as "dynamic dance," generating an abstract, swirling figure in real time. If a group of people hold up a sign reading "ocean," the AI instantly renders a photorealistic seascape. This feedback loop transforms passive viewers into active co-creators, echoing the participatory ethos of interactive art pioneers like Nam June Paik, but with the immediacy of modern machine learning.

The Raspberry Pi variant, selfusion-pi, is particularly noteworthy for its accessibility. By optimizing SDXL-Turbo’s lightweight inference engine and pairing it with YOLOv8’s efficient object detection, the developer achieved a fully functional AI art station that operates without cloud dependency. This makes the installation viable for schools, community centers, and developing regions where high-end hardware is scarce. The API version, meanwhile, is designed for larger installations and events, capable of handling multiple concurrent users and integrating with projection mapping systems or LED walls.

According to the developer’s Reddit post, the project was inspired by the growing trend of AI-driven public art and the desire to make generative models less like black boxes and more like collaborative tools. "I wanted people to see AI not as something that creates art for them, but something they create with," said r_giskard-reventlov in the comments. The installation has already sparked interest among educators and digital artists, with several universities expressing interest in replicating the setup for interactive media courses.

While the project does not yet include voice input or multi-user prompt voting—features that could further enhance group dynamics—it represents a significant leap in low-cost, real-time AI interactivity. Unlike static AI galleries that display pre-rendered images, this system thrives on spontaneity and collective input, making each viewing unique. Its open-source nature invites global collaboration: developers can add new detection models, customize prompt templates, or integrate it with AR headsets and IoT sensors.

As AI art continues to evolve from novelty to cultural institution, projects like this underscore a vital truth: the most compelling AI experiences aren’t those that impress with technical prowess alone, but those that invite human connection. With its elegant simplicity and profound interactivity, this installation may well become a blueprint for the next generation of public digital art.