Locate the object before assigning left or right
The person across a counter can occupy a different screen position after a camera change. A right-facing body, a right-turned face and eyes looking down at paper are different states. Locate the target in this view: the other person’s face, a blue circle on the table, or a door gap where a sound came from.
| Decision | A concrete description |
|---|---|
| Target | Qian’s face, or the accepted blue circle on the paper. |
| Location | Qian is screen right; the circle is lower on the counter between them. |
| Starting gaze | An looks at Qian before looking down at the paper. |
| Trigger | Qian indicates the circle before she shifts attention. |
For shot/reverse-shot, match directions and actual height relationships rather than making everyone eye-level. Recheck after a camera change instead of copying the previous view’s “left.”
Check the starting frame before revising the motion
The native video method treats accepted starts, ends, duration and dialogue as read-only boundaries. A real start frame looking into camera conflicts with a required start looking at the partner. Resolve the storyboard/keyframe rather than arbitrarily choosing one.
With a correct start, connect gaze movement to the accepted trigger: receive the indication, move attention to the target, describe supporting head movement, and settle. Preserve body, hands and objects that do not participate. Turning the head must not also turn the body, take the paper or relocate the person.
Static appearance may be omitted only when a real start image is actually supplied. With text alone, the copyable prompt still needs person, look, setting, composition and light. “As above” is not a reference frame.
Complete example: from Qian’s face to the circle on the counter
Accepted paper scene: An Ning is screen left and Qian is screen right and taller. A sheet lies flat between them on one counter. Its blue circle is a selected visual mark with no readable words. An initially looks at Qian’s face, then down at the circle after he indicates it. She does not know its meaning. No paper handoff, standing up or culprit discovery occurs. The author selects a fixed six-second two-shot; a real task must fit target capabilities. No start image exists here, so use a complete text-to-video brief.
A vertical, restrained live-action community-bookshop scene under warm ceiling light. The camera stays on the entrance side in a medium two-shot retaining both faces, hands and the paper on the counter. An Ning is an adult woman with short dark hair and a navy work shirt, screen left with her body facing Qian. Qian is an adult man with short grey hair and a pale-grey jacket, screen right facing her and visibly taller. The sheet lies flat between them, with its blue circle on the face visible to An and no readable words. An initially looks at Qian’s face, both hands resting on her own counter edge. Qian’s left index finger starts beside the sheet; his right hand stays on his own side of the counter.
Across six seconds, Qian moves his left index finger from the sheet edge to beside the blue circle and stops, leaving the sheet in place. After receiving this indication, An moves her gaze from his face to the circle, then slightly lowers her head and settles her attention there. Her body still faces Qian and both hands remain on the original edge. Qian keeps his position. At the end she looks at the circle without explaining it; the sheet is flat, the finger remains beside the circle and neither person has changed places. Fixed camera, no dialogue, and no added paper pickup, standing or gaze into camera.
These are selected example actions and visual facts, not an observation of an existing image or a generated video. For a real image-to-video start, verify the same placement, sheet and starting gaze before subtracting static details the image supplies. A face-only reference does not justify omitting the counter and hand states.
Check the candidate’s start, indication-triggered change and final gaze toward the circle rather than camera. Looking at it establishes changed attention, not reading, understanding a secret or discovering a culprit. Continue the next shot from this end instead of an unoccurred paper pickup.
Repair the located deviation and inspect the segment
For an incorrect final gaze, check target clarity and obstruction by hands or crop, then revise its location and endpoint. For wrong person-to-person direction, check this camera and blocking. “Natural” or “focused” cannot substitute for space.
Changes to accepted camera, placement or events belong in the storyboard rather than being silently substituted in motion text. A gaze repair that also alters face, clothes, custody or relationship meaning needs correction.
For a multi-shot conversation, see shot/reverse-shot continuity. For unchanging performance, see character reactions. This page first identifies the exact object of attention.
FAQ
Does a right-facing body imply rightward gaze?
No. Specify body, head and gaze separately. Looking at evidence on a table may shift attention while the body stays put.
Can motion text silently correct a wrong reference start?
Resolve the conflicting start first. The native method requires storyboard and keyframe starts to agree; motion text cannot simply ignore one.
Do it with the skill
Write gaze target, trigger and process within accepted starts and ends. Preserve body and custody and recheck screen direction after camera changes.
Read the method: Starting gaze, ordered motion and read-only endpoint · Eyeline directions and camera pairing for actual heights
About Drama SkillsThe short-drama-video-prompts skill on GitHub