The rise of diffusion models in AI image synthesis revolutionized computational image generation, challenging the traditional no- tion of photographic objectivity. This advancement posed a dual challenge for media artists: to harness AI’s potential while pre- serving artistic uniqueness. In Fencing Hallucination, we explored integrating AI models into a traditional hand-coded computation pipeline, introducing real-time interactivity to AI image synthe- sis and increasing the artistic control of the AI-generated images. The research culminated in an interactive installation exhibited at various venues, attracting over 800 audience visits. The system gen- erated authentic chronophotographs from diverse human-AI inter- actions while maintaining consistent aesthetic control. This project exemplifies the potential of combining AI models and hand-coded programs, which pave the way to better human-AI co-creative systems.