How to Make a Marketing Video: From Brief to a 9-Second Product Ad

Quick Answer

To make a marketing video, start with three decisions: who the video is for, what you want the viewer to do, and where the video will run.

Then turn that brief into a production workflow:

audience → marketing goal → channel → reusable assets → storyboard → generation → trim → QA → performance review

For this project, I created a short ad for a fictional foldable desk lamp called FoldLight. I prepared reusable product and scene references, planned the commercial as a timed storyboard, and generated a complete multi-shot sequence in Astorie with MiniMax H3.

The source generation was 15 seconds at 16:9. I kept the strongest first three beats and turned them into a tighter 9-second final cut.

Define Your Target Audience

A target audience should change what appears in the video.

It should not exist only as a line in a marketing brief.

For FoldLight, I defined the audience as:

People who study or work late at a compact desk and want focused task lighting without relying on harsh overhead light.

That decision gave me a clear visual problem to show.

Instead of placing the lamp in a generic studio, I built the ad around a nighttime desk. The environment included a laptop, notebook, pen, and a small work surface.

The product also needed to feel compact. That mattered because the audience problem was not simply "I need more light." It was "I need useful light in a limited workspace."

A simple audience brief can follow this structure:

For [audience] who struggle with [problem], show [visible benefit] in a situation they recognize.

For this project:

For people working late at small desks who dislike harsh overhead lighting, show how a compact lamp creates a more focused work area.

This gives the storyboard something concrete to prove.

Set One Marketing Goal

Before writing the storyboard, decide what the video is supposed to accomplish.

A video built for awareness does not need to do the same job as a video built for immediate conversion.

A simple way to separate them is:

I treated the FoldLight video as a consideration-focused product ad.

Its job was not to close a sale in nine seconds. The video did not show a price, discount, testimonial, or detailed purchase offer.

The more realistic goal was:

Get interested viewers to visit the product page and learn more.

A suitable CTA would be:

Explore FoldLight

This distinction matters because it changes what deserves screen time.

For a consideration ad, I cared more about making the product and its practical benefit easy to understand than about adding urgency or promotional copy.

A marketing video needs one clear job before it needs more shots.

Choose the Channel Before Production

Channel affects more than export size.

It can change:

• aspect ratio
• duration
• framing
• text-safe areas
• pacing
• CTA placement

For this project, I created one 16:9 master so I could keep the production workflow controlled.

I did not generate separate vertical and square versions during the test.

That does not mean the same 16:9 file should simply be cropped for every platform.

A vertical short-form version may need tighter framing and different safe areas. A landscape placement can preserve more of the environment. A product-page version may have room for more detailed proof.

The better rule is:

Adapt the creative for the placement instead of treating channel adaptation as a resize operation.

For this project, I wanted to solve one production problem first: could I build a clear product ad from reusable visual references and one structured storyboard?

How to Make a Marketing Video

Step 1: Prepare Reusable Product References

I did not start by asking the video model to invent the lamp from scratch.

First, I created a product reference pack.

It included:

• a three-quarter hero view
• a side view
• a top view
• a folded state
• an open, illuminated state

The hero image acted as the main identity anchor.

The other views supplied structural information that was harder to read from one angle. The folded and illuminated references defined two important product states that would later appear in motion.

In Astorie, I saved the approved images as reusable Elements so I could reference them again later.

The purpose of the pack was not to guarantee perfect consistency. It was to give the later generation stronger visual constraints.

That distinction matters.

A product can remain visually consistent while still reproducing the wrong hinge, wrong proportions, or an invented detail. Consistency is not the same as accuracy.

Step 2: Prepare the Scene References

Next, I built a small scene pack.

I used three main references:

• a primary nighttime desk scene
• a top-down view of the same desk
• an improved illuminated version of the workspace

The first image established the environment.

The top-down reference made the desk layout easier to understand. It showed where the laptop, notebook, lamp, and other objects sat relative to one another.

The illuminated version gave the later video a visual target for the benefit state.

This was more useful than collecting several mood images with unrelated layouts.

A scene reference should define spatial logic, not just atmosphere.

Step 3: Turn the Message Into a Timed Storyboard

The next step was to decide what each part of the ad needed to accomplish.

I initially planned a longer sequence, but the strongest final cut used three beats:

The storyboard also defined camera language.

The opening used a wider view of the desk. The second beat moved attention toward the product. The third returned to the workspace so the viewer could see what changed after the lamp was activated.

This is why I find it useful to give every shot a marketing job.

A shot should not exist only because it looks cinematic.

For each storyboard beat, I wanted to know:

• what the viewer sees
• how long the beat lasts
• what the camera is doing
• what changes during the shot
• what the viewer should understand before the next beat begins

That makes the storyboard more useful than a simple row of pretty frames.

Step 4: Reference the Approved Assets in One Structured Prompt

Once the assets and storyboard were approved, I brought them into one structured generation prompt.

I used @ references for the product images, scene images, and storyboard.

Each reference had a specific job.

The main product image established identity. The side and top views added structural context. The folded and illuminated references showed product states.

The scene references controlled the environment and layout. The storyboard carried the timing and narrative sequence.

I did not add references just because more references were available.

The useful question was:

What decision is this reference helping the generation make?

That kept the prompt easier to reason about.

Step 5: Generate the Full Sequence

I generated the commercial as one complete multi-shot sequence.

I did not create three separate clips and stitch them together afterward.

The source version used:

• MiniMax H3
• 16:9
• 15 seconds

The prompt described the full sequence with timed beats.

That let the production stay centered on the commercial as one piece rather than treating every shot as an isolated test.

The first part of the result worked especially well.

The opening showed the full nighttime desk from above. The generated video also introduced a brief light sweep moving across the desk from left to right.

I had not treated that light sweep as a required control in the storyboard. It appeared in the result, but it gave the opening more visual energy, so I kept it.

The second beat moved toward the lamp as it unfolded and activated.

The third beat returned to the wider workspace. The lamp was now open and illuminated, and the warmer working area made the benefit visible without explanatory text.

Step 6: Trim the Strongest 9-Second Cut

I did not keep all 15 seconds simply because 15 seconds had been generated.

The strongest version of the ad ended after the first three beats.

That produced a final cut of about 9 seconds:

0–2.5s — Hook

A high view establishes the nighttime desk and the folded lamp. The brief light sweep gives the opening an additional visual hook.

2.5–5.5s — Product Reveal

The framing moves closer to FoldLight. The lamp unfolds and switches on.

5.5–9.0s — Benefit

The wider workspace returns. FoldLight is fully open, and the warm task light now covers the main working area.

The three shots form a complete message:

problem → solution → visible benefit

I preferred that to keeping extra footage simply to reach the original duration.

This became one of the most useful lessons from the project:

A longer source generation does not require a longer final ad.

Generation gives you material. Editing decides which material deserves to stay.

Step 7: Run Creative QA

Before publishing, I checked the video as a marketing asset rather than asking only whether it looked good.

First, I checked product accuracy.

The lamp needed to remain recognizable across the folded, unfolding, and illuminated states. The hinge, base, light bar, color, and overall proportions all mattered.

Then I checked the environment.

The desk, laptop, notebook, and room needed to feel like the same workspace. The lighting change also needed to make sense.

Finally, I checked the story without relying on captions.

Could someone understand that the first shot showed the problem?

Did the second shot clearly introduce the product?

Did the third shot show a visible improvement?

If those three ideas are not readable from the visuals, adding more polish will not solve the underlying problem.

How to Evaluate Marketing Video Performance

Creative QA and marketing performance are different questions.

Creative QA asks whether the video was made correctly.

Performance evaluation asks whether it achieved its marketing goal.

For FoldLight, the goal was to drive interested viewers toward the product page.

That makes product-page traffic or click-through rate more relevant than raw views alone.

I would separate the metrics into three groups.

Attention metrics

These help diagnose whether the creative earns and holds attention.

Examples include:

• early retention
• watch time
• completion rate

If viewers leave almost immediately, the opening may be worth revisiting.

That does not prove the hook is the only problem, but it gives you a useful place to investigate.

Action metrics

For this campaign goal, the main action is a product-page visit.

Useful metrics include:

• click-through rate
• product-page visits

A video can have solid completion and still produce weak clicks.

That may suggest the creative is watchable but does not create enough product interest or a strong enough next action.

Downstream metrics

Clicks are not the end of the funnel.

I would also look at:

• landing-page engagement
• add-to-cart activity
• conversion rate

If the video produces clicks but downstream performance is weak, I would not immediately blame the video.

The traffic may be poorly qualified, the message may create the wrong expectation, or the landing page itself may need work.

Performance data should guide the next test, not produce instant causal conclusions.

Common Marketing Video Mistakes

Starting With a Prompt Instead of a Marketing Decision

A detailed prompt cannot fix a vague campaign.

Define the audience, goal, and channel first. Then decide what the video needs to show.

Trying to Show Too Many Benefits

A short video becomes harder to read when every feature competes for attention.

The FoldLight ad focused on one visible benefit: creating a focused working area at night.

That was enough for the project.

Reusing One Format Everywhere

A finished 16:9 video is not automatically ready for every placement.

Check framing, ratio, pacing, text-safe areas, and CTA before adapting the same creative to another channel.

Judging Success by Views Alone

Views tell you that the video was played.

They do not tell you whether the right people understood the product or took the intended action.

Measure success against the original campaign goal.

FAQ

How long should a marketing video be?

There is no single ideal length for every campaign.

The video should be long enough to complete its marketing job without keeping unnecessary beats.

In this project, I generated a 15-second source sequence but found that the strongest finished version was about 9 seconds.

Can AI generate a complete marketing video from one prompt?

A full sequence can be generated from one structured prompt, but the prompt is only one part of the workflow.

This project also relied on approved product references, scene references, a timed storyboard, creative QA, and trimming.

The final result came from the full production process, not from a prompt alone.

Should I make a different video for every platform?

Not always.

You can reuse the same campaign idea and visual assets, but each placement may need different framing, aspect ratio, pacing, safe areas, or CTA treatment.

Adapt the creative to the channel instead of assuming one finished file can simply be resized everywhere.