Amidst the recent paradigm shift in the film industry driven by generative AI technology, this study proposes and analyzes a production workflow for ‘hybrid films’ that combine live-action footage with AI-generated video, verifying its effectiveness through the actual production case of the short film <JOI>. This research centers on a production method that cross-edits live-action footage, which depicts objective reality, with generative AI video that expresses the subjective inner world of the AI protagonist. Specifically, <JOI> applied AI in four domains, including generating weather conditions linked to the audience’s IP address and creating animal performances that were impossible to direct. The results showed that the proposed five-step workflow achieved significant success in terms of expanding creative expression and production efficiency. However, it also revealed clear technical limitations, such as maintaining visual consistency between live-action footage and AI-generated results, and the uncertainty of controlling the AI. Ultimately, the significance of this study lies in its contribution of a practical production process as research data for future creators, and in its exploration of the potential for a new cinematic aesthetic that visualizes the psychology of a non-human subject through AI.