What should motion data remember?
A working thesis on temporal semantic labeling: movement becomes more useful when it carries intention, phase, contact and story context — not only transforms over time.
Experiments, production lessons and practical research from the work behind the work.
QUESTION → TEST FAILURE → FINDINGProduction problem. Hypothesis. Test. Result. Limitation. Current status. Publish when the work leaves behind something the next production can use.
These are active lines of inquiry, not finished claims. A note becomes evidence only after the work has been tested.
A working thesis on temporal semantic labeling: movement becomes more useful when it carries intention, phase, contact and story context — not only transforms over time.
Prompting can describe an outcome. Direction shapes why it happens, what changes inside the performer and which imperfection makes the action believable.
A transparent production log exploring where Blender, Unreal and controlled AI assistance genuinely compress a 3D pipeline — and where senior craft still refuses compression.
Tests around crowds, variation and motion systems that preserve intention instead of multiplying randomized bodies.
A position on the work that becomes more valuable when execution accelerates: taste, story judgment, performance direction and ownership of the final choice.
Using Before Frame One to show how a compact story grammar can reduce visual drift, revision loops and accidental worldbuilding.