The lid is still on the face-cream jar when the hand reaches it. A second later, the jar is open and the lid is resting beside it—or already in the person’s hand.
Both images make sense on their own. A closed jar is a normal starting point. An open jar with the lid removed is a normal ending. Yet the action between them feels strangely empty. The hand touched the lid, but it never seemed to unscrew it.
This is a hypothetical scene, but it describes a useful problem to watch for in AI product clips. The result arrives without the small physical events that should have produced it.
A lid has more than two states
“Closed” and “open” are easy to describe. Opening is harder because it is a chain of changes.
The fingers first make contact with the lid. They tighten around its edge. The lid begins to rotate while the jar stays on the table. The seal loosens. A narrow gap appears between lid and rim. Only then does the lid lift away, followed by a short pause that lets the viewer register the new state.
If the clip skips rotation or separation, the lid appears to teleport. The beginning and ending may still look attractive, but they no longer explain each other.
The same missing middle appears elsewhere: a jam-jar cap never turns, an earbud reaches a hand without leaving its slot, or a gift-box ribbon falls away before the knot loosens. A detergent cap may lift without clearing the threads.
These examples share one thing: contact should lead to movement, movement should lead to release, and release should lead to the final position.
Give the middle room to exist
For the cream jar, the action can be made smaller before any video is generated. Instead of asking for a hand to reach in, open the jar, scoop out cream and apply it, stop after the lid comes off.
Plan three parts: the fingers settle while the jar stays still, the lid rotates until a gap appears, and the hand lifts it beside the open jar.
The contact point needs to stay visible. A high angle may show the fingers but hide the gap under the lid. A very low angle may show the rim while the hand covers the rotation. A three-quarter view can often keep the fingertips, lid edge and jar opening in the same frame.
Quiet the rest of the shot. Reflective packaging, moving labels, hand jewelry and a camera orbit compete with the lid. A steady camera and simple background keep the action readable.
This matters when planning an AI product video from a single image. One product photo may show the shape and finish clearly, but it may not reveal the thread under the lid or the gap that appears during opening. An additional open-jar reference, photographed from a similar angle, can provide information the closed image does not contain.
If one shot still asks too much, divide it at a natural pause. End the first clip after the lid loosens, then begin the second with the fingers holding it. Preserve the jar position, hand placement and movement direction across the cut.
Check the change, not just the endpoints
Once the action and references are prepared, Vidu Q3 can be used to make a short draft from text, an image or visual references, with prompts focused on the hand, lid and camera. Available duration, resolution, motion and audio options depend on the selected model and workflow, so choose only the settings needed to judge this small interaction.
The first review should happen at normal speed. Does the opening feel like one continuous action, or does the eye notice a jump? The second review should move frame by frame through the contact, first rotation, visible gap and lift.
Look for simple cause-and-effect clues. The fingers should move with the lid rather than slide across it. The lid should rotate before it rises. The gap should widen as the seal releases. The jar should not shift unless the hand applies enough force to move it. After separation, the lid should remain somewhere in the scene instead of vanishing.
If only the separation fails, that short transition is the part to revise. There is no need to cover it with a faster cut or add more decoration. Hiding the missing movement does not restore the cause.
Recognizable packaging or real people also require permission before their images, voices or likenesses are uploaded and altered. A generated opening sequence can illustrate how an action might look, but it should not be presented as evidence of a product’s real seal, texture or performance.
Two correct pictures are not enough to make an action believable. What earns the viewer’s trust is seeing one state become the next.