The Main Difference
Before generating, it is easy to write one large instruction: “make a beautiful cinematic video with a product, lighting, camera motion, and no mistakes.” It contains everything except a clear task. The model has to stage a photograph, shoot a film, invent the edit, and remember the object at once. You may get lucky. A working method is not built on “maybe we get lucky.”
An image prompt answers one question: what must be visible in one still frame? A video prompt answers another: what changes over time, and which visible state ends that change?
The shot-plan lesson can help on a real project, but it is not required for the exercise below. One simple scene is enough here.
Image Prompt
Imagine the scene has frozen and you can take only one photograph. Describe the main subject and its state, location, composition, shot size, angle, lighting, materials, what must be preserved, and what must not be added. Duration, camera movement, and a chain of events do not belong here.
Subject and state:
Location:
Composition and angle:
Lighting and materials:
What to preserve:
What not to add:Video Prompt
Once the starting state is clear, describe one readable change: the start, main action, one camera job only if it helps, what stays unchanged, the end, and the main risk. The end is not poetry; it shows whether the action finished.
Start:
Main action:
Camera:
What to preserve:
End:
What must not break:If the action does not fit the selected duration, simplify it or split it at a meaningful boundary. Speeding up everything is not an editing solution. It is trying to push a sofa sideways into an elevator.
One Example: Bottle and Drop
Suppose you need an advertising frame of a glass bottle on black water, followed by one falling drop. It is one world, but two different instructions.
First, the Image
This is a prompt to copy into an image generator. It builds the starting frame: nothing falls, and the camera goes nowhere.
A glass bottle stands upright on glossy black water. Clean silhouette, a slightly angled overhead view, cool soft backlight, and thin reflections along the edges. Preserve the bottle shape and empty space on the right. Do not add a second bottle, accidental text, petals, or bright colors.Then, the Video
Now use a separate prompt for motion. It does not change the scene world: the bottle, lighting, and background remain the same; only the water beside it changes.
One drop falls into the water beside the stationary bottle. One soft ripple spreads across the surface; the camera stays stable and moves barely closer. Preserve the bottle shape, cool lighting, and clean background. At the end, one ripple is fully visible to the right of the bottle. Do not add text or extra objects.Weak and Revised Versions
A weak request asks for “a premium cinematic video with a bottle, water, a drop, beautiful camera work, a logo, and atmosphere, with no mistakes.” It does not say what matters most, how the scene begins, what changes, or where the action ends. “Beautiful” does not control a camera, and “no mistakes” does not name the mistake to prevent.
The revised route separates the task: first, one bottle on black water with cool side lighting and a clean background; then, one drop beside it, one ripple, a stationary bottle, a visible ending on the right, and a rule not to change its shape or add text.
Practice and Check
Choose a simple scene and write two 2–4-line texts: a still frame and a change to that frame over time. You do not need to run a generation to complete this exercise; you are testing the difference between the instructions now.
Where to Go Next
You can next prepare and review a static production frame, choose a real generation route, or build a controlled video segment. These are recommendations, not a required ladder.
You have already completed this page’s independent outcome: you can distinguish a frame description from a movement instruction.