HUMBLE. (Recast)
Personal study · AI video and motion · 2026
Each scene was recast from a single reference sheet of me in one consistent look (black tee, black cap) using Higgsfield's Genjutsu, which swaps a person in existing footage while keeping their original performance. Then I pushed the image somewhere new: a painterly, cel-shaded grade with warm amber light against deep violet shadow. That turns three very different locations into one visual world.
What I was testing
Holding one identity across three very different shots: tight close-ups, a wide tableau, and a face that shrinks to a few pixels as the camera cranes away. Keeping the original performance intact: the hand-to-mouth gesture, the head drops, the timing, all inherited from the source footage. Unifying mismatched footage through grade and style so it reads as one piece, not three experiments.
Tools: Higgsfield (Genjutsu), ffmpeg for frame analysis and QC, [Premiere / After Effects for edit and grade].
This is a non-commercial personal study.
Frames
↑Back to Top