DRIFT

Personal AI-driven motion experiment on worldbuilding and continuity.



Initially framed as an open R&D test, this project evolved into a focused exploration of AI workflows — combining generative tools with cinematic motion logic. The aim was to test how far consistency, structure, and visual control could be pushed across a short-form video.

Concept

Beyond the prompt: The project questioned whether prompt engineering has peaked — shifting focus toward creative direction, continuity, and structure. With emerging models behaving more like production tools than black boxes, the experiment reframed AI as a medium for directed storytelling rather than random generation.

Design

Structured flow: Explored cinematic continuity across shots — including consistent characters, environments, styling, and guided camera choreography. Built a short, scored sequence using multi-model workflows (GPT-4o, Higgsfield, Runway Gen-3, diffusion) to merge spontaneity with directorial intent.

Delivery

Motion as identity: 
Developed and tested a scalable creative pipeline across image generation, motion interpolation, editing, and audio. The result: a structured visual test prioritizing consistency, pacing, and motion language over standalone visuals.