This piece explores prompt‑driven image and motion design. Starting from a blank canvas in Midjourney, I built a character that feels half ritual, half glitch — a figure whose head dissolves into smoke while the body stays sharply lit and grounded. The two stills on this page are the beginning and end frames of the sequence, designed as anchor points for the movement.
To get there, I iterated prompts beyond simple “style” tags: combining lighting instructions (studio backlight, directional haze), wardrobe and posture details, and constraints on composition so the model would generate frames that could sit in a coherent timeline rather than as isolated images. Once the visual language was stable, I produced a set of variations and stitched them into a short video edit in CapCut, working with speed ramps and cuts to keep the motion as tense and minimal as the stills.
For me, prompt engineering here is less about tricks and more about control — using language to steer Midjourney toward a specific atmosphere, then treating the outputs as raw material for a directed sequence rather than finished artwork. This project shows how I translate abstract narrative and mood into repeatable prompts, and how I bridge AI‑generated frames with traditional video editing to arrive at a cohesive piece.