Alibaba · model workflow guide

Wan 3.0 First & Last Frame Guide

First and last frame control is most useful when the frames are treated as production anchors: the first establishes the exact starting state and the last defines where the shot should resolve.

How GenVideoKit structures Wan 3.0 prompts

Describe composition, subject motion, environment motion, camera behavior and temporal continuity explicitly. Keep cause-and-effect clear, maintain stable identity/geometry, and specify reference relationships when used. Audio direction may be included for current multimodal Wan workflows.

Text → videoImage → videoVideo → videoNative audioReferencesFirst / last frame

Production shortcut

Treat the first frame as the exact start state and the last frame as the intended resolution. Keep identity, screen direction and environment consistent between them.

Reference pricing snapshot
$0.1 / second

First frame

Use a clean starting state with readable subject pose, environment and screen direction. Avoid an ambiguous pose that already looks mid-transition unless that is intentional.

Last frame

Define a stable end state that can also become the next clip’s reference. The cleaner the final subject pose and environment, the easier it is to continue the sequence.