Video Generation Prompts: 8 Examples for Short Product Videos

You want to bring a product photo to life. The model turns the camera and the product with it; a few seconds later the packaging looks like something else. Adding more “cinematic” to the prompt does not describe the problem. You first need to choose what will move.

This guide contains eight video-generation prompts for product intros, Reels openers, workshop footage and a short brand film. Some examples are text-to-video, others image-to-video. Each is planned as a single shot; in the multi-shot example we generate clips separately and join them in the edit.

How do you write a video-generation prompt?

A video-generation prompt describes what is in the shot and what changes over time. Camera movement, the subject’s action and sound are separate decisions. In a short shot, choosing one of them as the main movement is a good start.

First complete this sentence: “In this clip, the viewer will see [one action] by [the subject].” If you cannot complete it, narrow the idea before writing a longer prompt.

Use: [product intro, Reels opener, background video].
Input: [text only or a start frame].
Scene: [subject and location].
Action: [one main movement].
Camera: [framing and one movement].
End frame: [where the shot finishes].
Sound: [ambience, effect or dialogue if supported].
Preserve: [product, person or scene details].

Set duration, aspect ratio and resolution in the tool’s supported settings. The durations below are editing targets; they do not mean every tool supports the same length. For the first attempt, use the short clip your tool supports.

The difference between text-to-video and image-to-video

Which method you choose depends on whether you have a start frame and on what should change. In text-to-video you describe the whole scene. In image-to-video the frame already shows the product’s colour, texture and placement; the prompt’s job is to describe motion and the end frame.

Google’s image-to-video guidance recommends describing the motion rather than re-describing the start frame at length. If the uploaded image already shows the product’s colour and the setting, do not spend most of the prompt repeating them.

When you animate a photo, the product often changes because unseen surfaces are regenerated. As the camera pushes in, the back, interior or label of the product is reinterpreted by the model. A sharper start frame and a smaller movement reduce that risk; they do not remove it.

Eight practical examples

Each example solves one job. Replace the brackets with your own product and delete the sentences you do not need. On a tool without sound support, leaving the sound line in the prompt does not create a new capability; it only makes the text longer.

1. Push the camera in slowly while the product stays still

A calm opening that lets the product’s shape be examined is easier to judge than a complex rotation. Input: an approved start frame showing the whole product. Method: image to video. Edit target: one 5–8 second shot; 9:16.

The camera moves very slowly toward the product in the start frame. The product stays in place and in the same orientation; the camera does the moving. At the end of the push-in, the top and bottom edges of the product are still inside the frame. The original light and background are preserved. The camera comes to a soft stop near the end of the shot.
If sound is supported, use only light room ambience; do not add speech or music.

Check: as the product grows, do the cap, stitching or print change? If so, reduce the distance of the camera move. This instruction is not a true physics simulation or a guarantee of exact product preservation.

2. Animate only the environment when turning a photo into video

A scene does not need the product to lift into the air to feel alive. A small environmental movement can be enough for a calm brand language. Input: a photo of a tote bag near a window. Method: image to video. Edit target: 5–8 seconds; 9:16.

The camera stays completely still. The bag and chair in the start frame do not move. The thin curtain in the background ripples slowly once in a light breeze and settles. The direction of light, the bag’s print and the scene layout are preserved. Keep the end frame close to the opening composition.
If sound is supported, include a faint fabric rustle and low room ambience.

Intended output: a lightly moving scene that preserves the product’s readability. We do not promise a seamless loop; if you need one, check the first and last frames in the edit.

3. One material detail for a Reels opener

Choose the detail you will show in the first second. One clip does not have to carry the product reveal, the use and the result. Input: none; use a reference if real product accuracy matters. Method: text to video. Edit target: 4–6 seconds; 9:16.

A close-up of dark blue linen fabric resting on a light-coloured worktable. An adult hand slowly lifts one corner of the fabric; the natural fold and woven texture become visible. The hand completes the movement and stops. The camera is static and level with the fabric. Soft daylight comes from the left. The frame is clean; there is no text in the image.
If sound is supported, only a short fabric rustle is heard.

Check: inspect the contact between hand and fabric and the fingers. If something is wrong, simplify the hand movement; do not change the camera in the same attempt. This is an opening idea, not a guarantee of views or conversions.

4. A short workshop shot for a local business

Instead of a camera flying in from outside, focus on a small piece of work happening inside. Input: none; if it will represent a real business, use your own venue reference. Edit target: 6–8 seconds; 16:9 or 9:16.

A close-up of a workbench in a small ceramics studio. An adult craftsperson smooths the rim of an unglazed bowl with a damp sponge in one slow movement. The camera is static; the hands, sponge and whole bowl are visible. In the background, a few plain shelves are out of focus. Window light reveals the clay’s moisture and surface texture.
If sound is supported, include the faint friction of the sponge and calm studio ambience. No dialogue.

Check: do not present a workshop or production process that does not exist as if it were real footage of the business. If it is used as a concept video, label it in the appropriate context.

5. A short demo shot with a presenter

Loading a long speech, hand movement and product interaction into one clip creates many checkpoints. A single short line is enough for a first draft. Input: product reference; an approved character reference if the tool supports it. Method: reference-guided generation. Edit target: 6–8 seconds; 9:16.

An adult presenter sits at a plain desk with the notebook from the reference open in front of them. The camera is at eye level in a static medium shot. The presenter slowly turns one page and stops their hands. The notebook’s size, cover colour and binding should match the reference. Natural window light, a relaxed delivery, a plain background.
If speech generation is supported, the presenter says in Turkish: Bu sayfada haftanın işlerini birlikte görebilirsiniz.

Check: Turkish audio, lip movement and product details are checked separately. If the tool does not support speech, generate a silent clip and add the voice afterwards. This is a demo line; it is not an invented customer experience or testimonial.

6. A background video that does not compete with the page

A background video’s job is not to compete with the headline. Think about the intensity of motion together with text readability. Input: none. Method: text to video. Edit target: a short silent clip; 16:9.

Two translucent blue paper layers rest on the right side of an off-white surface. Soft light passes through the layers. Only the edge of the thin sheet on the right moves very slightly; the rest of the scene is still. The camera does not move. The left half stays empty, plain and at the same brightness. No text, logo or new object appears. Silent footage.

Check: place the actual headline over it and test readability. Ship a small file and a still alternative; looking good does not justify a heavy page. Users who prefer reduced motion can be served the still image.

7. Build a short brand film from three shots

Here we do not generate the whole film from one prompt. We show the same tote bag in three separate shots and join the clips in the edit. Target total length is about 12–18 seconds; choose the usable sections by cutting. Shared inputs: approved references of the same bag, the same light direction and colour palette. Product scale and print should be checked between shots.

A close-up of the stitching on the tote bag in the reference. The bag stays completely still. The camera slides a very short distance to the right along the seam and stops. Soft side light, natural fabric texture. Do not change the print or stitch pattern. Do not add text.

The bag from the reference lies open on a wooden table. An adult hand enters the frame, places one closed notebook inside the bag and withdraws. The camera is static in a medium close-up; the whole bag is visible. The same soft light and light surface are preserved.

The bag from the reference is fully visible on a chair. The camera pulls back very slowly and stops. The bag stays on the right; on the left a plain area opens where a short headline can be added in the edit. Do not add new objects, generated logos or text.

Edit note: use the section of each clip that stays intact. Add the logo and closing message after generation. If the product looks different in the first and last shots, fix consistency rather than hiding it with a transition.

8. Fix a failed shot with one variable

This text is for a tool that can edit the previous video. If the tool only regenerates, apply the fix as a new attempt with the start image and the full prompt.

In this shot, change only the camera movement. Remove the fast push-in; keep the camera static for the whole shot. Preserve the product’s position and appearance, the light, the background and the action in the scene. Do not add a new shot, transition, person or text.

Check: do not delete the previous version. Pause both videos at the same point and compare. Details you asked to preserve may have changed during the edit too.

Camera movements and common terms

Camera terms are decisions, not decoration. If you ask for a static shot and also write “dolly in”, you have made two different movement requests. Choose one movement per shot and state where it starts and stops.

Sound is a separate decision too. If dialogue, effects and music are requested in the same line, it stays unclear which should lead. Choose one sound layer in the first attempt; music is often more controllable in the edit.

How should you adapt the prompt when you change models?

Labels and structures that work in one model may not mean the same in another. When you move to a new tool, read its own guide first, then write the scene as a flowing single-shot paragraph. Instead of pasting tables into the generation field, use the prompt for the relevant shot.

LTX’s official guide explains that tags and shot structures written for other models should not be carried over directly, and recommends a flowing paragraph for a single shot. Read the prompts here shot by shot, not as a table.

Example outputs from the source projects

The demos here were not generated with the original prompts on this page. The aim is to study motion from a single frame and camera movement in the sources’ own examples. Instead of copying third-party media onto our server, we link to the source; licensing and attribution terms can differ per asset.

Lightricks official GitHub demo

In the image-to-video examples in the LTX-Video repository, study where motion generated from a single frame breaks down. The first animated GIF in the repo is a demo of the older LTX-Video; it should not be presented as a current test. The repository points to the LTX-2 project for ongoing development.

The demo media links were verified in the source repository; the animation frames and original generation settings were not independently examined for this piece. No performance score or reproducibility claim is drawn from demo quality.

Google example: separating the start frame from the motion instruction

The blue pickup and sunflower field image in Google’s video guide is not a video output; it is the start frame used by the source. Google’s example describes a rising camera movement starting from that frame. Apply the same logic to your product: show the frame, write the motion separately.

The compass push-in example in Google’s camera-movements section shows a wide frame narrowing onto one detail. When adapting it to your product, try the camera movement alone rather than adding a product rotation at the same time.

Plan your first attempt

Fill in these fields before generating. The goal is not a complete storyboard but knowing what to check in the first clip:

• Placement — Where will the video run: site, Reels, ads?

• Main movement — What is the one main movement: the subject or the camera?

• Start image — Do I need an approved start frame?

• End frame — What should be visible when the shot ends?

• Sound — Will sound be added during generation or in the edit?

• Preserve — Which three details must not change to call it a success?

If you do not have a start image, use the video start-frame example in the image-generation prompts guide. Save the approved frame, then generate the first clip with a single movement.

Frequently asked questions about video prompts