Move the Scene Without Losing the Frame That Approved It

Use one image as the opening frame, or add a second image to define both the first and last frames. Standard Wan 3.0 can generate 2–30 seconds at 480P, 720P, or 1080P with optional audio. The frame controls the visual state; the prompt should describe motion, camera behavior, timing, and sound rather than restating every pixel.

Choose a clear focal subject, readable edges, intentional lighting, and enough surrounding space for the requested movement or camera travel. Avoid accidental blur, heavy compression, unclear cutouts, conflicting light directions, and crops that hide the space the action needs. Remove tiny overlays unless preserving them is essential and you can review them closely.

Add a last frame when the clip must reach a defined visual destination: a product reveal, transformation, transition, pose, lighting state, or edit point. Make sure the first and last frames can plausibly connect through the chosen duration. Use only a first frame when the opening identity matters but Wan 3.0 may decide the ending composition.

Name the source details that must remain recognizable, then state what moves first, how far it moves, how quickly it develops, how the camera responds, and where the shot ends. Add fabric, smoke, water, foliage, reflections, or other environmental motion only when it supports the subject. Put sound cues on the same timeline as the visible action.

Start from an image that clearly shows silhouette, materials, label placement, and viewing angle. State which attributes must remain stable, then request one controlled change such as a turntable reveal, lighting sweep, camera arc, or restrained environmental motion. Review geometry, text, logos, and reflections before judging decorative effects.

Use a front or three-quarter view with clear facial features, natural posture, and limited occlusion. Ask for observable motion such as a gaze shift, small head turn, breathing, hair movement, or a controlled camera push. Keep the first test restrained, then inspect eyes, mouth, hands, hair edges, and face shape at the opening, midpoint, and ending.

Match the action to what the frame can plausibly support. A close portrait suits expression, breathing, hair movement, or a gentle push-in; a wider environment can support parallax and larger camera travel. If the action requires unseen poses, new scenery, or a demonstrated performance, simplify the move or use Reference to Video with stronger motion guidance.

Use a short duration to validate identity and one motion path, then extend toward 30 seconds only when the scene needs more development. Choose 480P for lightweight tests, 720P for clearer review, or 1080P for closer detail inspection. Turn on audio when the prompt includes dialogue, ambience, music, or an effect that must align with the motion.

Stay in Image to Video when one opening frame or a first-and-last-frame pair contains the main visual decision. Move to Reference to Video when separate images should define identity or style, a video clip should demonstrate movement, or audio should provide timing. Give every added source one job instead of using a large undifferentiated reference set.

Compare the opening frame with the source, then check identity or product geometry at the midpoint and ending. Review motion path, camera stability, background deformation, edge quality, lighting continuity, visible text, last-frame accuracy, and audio timing. Record one correction for the next pass and keep the strongest result with its prompt and settings.