Prompt craft · 6 minute read
Why camera verbs need endpoints
“The camera pushes in” tells the model how to move. It does not say what the move is for.
A useful camera direction has three parts: starting frame, motion, and ending frame. “Slow push-in from a medium shot to a close detail of the repaired seam” is easier to visualize and easier to judge than “cinematic camera movement.”
The endpoint creates priority
Every move reallocates attention. An orbit reveals form. A push-in isolates detail or emotion. A pull-back reveals context. Tracking preserves spatial relationship while the subject moves. Without an endpoint, the model may continue moving after the important reveal, overshoot the subject, or invent a new framing goal.
Write the move as a test
Ask whether the final frame can be checked without interpreting mood. “End with the latch centered and fully closed” passes. “End on a powerful feeling” does not. The first gives you a concrete reason to keep or reject a render.
Start in a medium side view of the mechanic's hands. Track left at workbench height as she closes the brass latch, ending on a tight detail with the latch centered and both hands still.One move is usually enough
Camera stacks often contradict themselves: “drone shot, handheld push-in, orbit, rapid zoom.” Choose the move that reveals the most important change. If the subject action already creates energy, let the camera behave more simply.
Use motion to reveal, not decorate
Before adding a camera move, finish this sentence: “The camera moves so the viewer can see…” If you cannot complete it with a concrete subject, detail, relationship, or environment, a static composition may be stronger.
Apply the method in the prompt template builder, or return to the full Seedance 2.5 tutorial.