Map multiple Genjutsu characters to ordered references

A group shot needs a casting map before a style description. Identify the subjects in the opening frame, then connect each one to a numbered reference. Use stable descriptions such as “leftmost dancer at the start” rather than only “the person on the left,” whose position may change during the shot.

Make a small cast table

Source role	Reference	What the instruction identifies
Leftmost performer at the opening	Image 1	Identity and appearance
Central performer at the opening	Image 2	Identity and appearance
Rightmost performer at the opening	Image 3	Identity and appearance
Intended environment	Image 4	Scene only

Keep the uploaded order aligned with this table. If you replace image 2, its role remains number 2. If you remove it, revise the later numbers before submitting.

Describe a crossing without renaming people

An original starting prompt:

Cast the leftmost opening performer from image 1, the central opening performer from image 2 and the rightmost opening performer from image 3. Follow each performer’s movement through the crossing. Use image 4 for the setting. Keep the source camera direction.

This is an untested editorial suggestion. It describes the intended correspondence; it cannot guarantee that a model tracks identity through every occlusion.

Inspect three moments

Review the first clear frame, the point where subjects overlap, and the first clear frame after they separate. Check whether identity, costume and relative scale remain consistent. A pleasing opening frame can hide an identity swap later in the clip.

For a difficult crossing, begin with a shorter useful section that still meets the input-duration requirement. Keep the source and references constant while clarifying one role. Record the revision rather than replacing the previous attempt.

Learn from World Shift without hiding its limits

The archived official World Shift workflow contains more references than the public API accepts. The library preserves its original inputs for inspection and offers a separately labeled seven-image adaptation: six character roles and one setting. This is a EveryGen AI adaptation, not an independently verified reproduction.

Use the preset library to compare the original with its adaptation. Confirm that all seven references load before trying the adapted brief in Motion Transfer. The same method applies to simpler pairs or three-person scenes: give every input a role, then review where identities are hardest to follow.

See reference preparation for choosing useful views. This guide proposes a workflow for the current site; it does not claim access to undisclosed settings from the original creator.