AI Talk: Creating a Marketing Video with AI — in Two Hours Flat

I have never created a marketing video. And certainly not one in two hours flat. But I just managed to do that this week.

The problem I was trying to solve was related to a charity event we are planning to host in a few weeks. We wanted a crisp marketing message that could be promoted on social media. I don't mean LinkedIn.

As is usually the case—at least with me—the thought was: let's see what I can do with AI.

This blog is about what I managed to do in a couple of hours, and how.

The Context

The Hindu Religious and Cultural Center (HRCC) is located in Morgantown, West Virginia. I am one of the members of its Humanitarian Committee, which organizes charity events.

The event we were working on was the Gandhi Day Walkathon & India Festival.

So here is what I typed into the ChatGPT prompt window:

We are trying to create a short video montage to promote GDW. I want to include the following aspects: 1) Honor Mahatma Gandhi's message of peace and non-violence. 2) The focus on the walk—a take on the Salt March organized by Gandhi. 3) Our focus on local charities for the past two decades. 4) Celebration of Indian culture—which has taken two forms: a) a variety of Indian foods sold with the proceeds going to charity, and b) a cultural program involving kids and adults. I have a collection of images of the walk, food, and cultural programs over the years. What is the best way to create this video?

ChatGPT's response:

I would make this as a 60–75 second story-driven montage, rather than a chronological slideshow of past events. The strongest narrative is: Gandhi's message → walking together → two decades of giving → celebration of Indian culture → this year's invitation.

The Process

I liked the approach, so I started uploading a few images.

Then I realized I had hundreds of images, and it seemed crazy to take the time to upload them one at a time. So I asked my ever-ready AI companion what to do.

Zip them up and upload them, came the response.

So that's what I did. I zipped multiple folders of photos and uploaded them.

ChatGPT then came up with a detailed outline of how it would structure the video montage:

Time	Theme	Visuals	On-screen message
0–8 sec	Gandhi	Gandhi portrait / Salt March image / Gandhi artwork from our flyer	“Honoring Mahatma Gandhi's message of peace, non-violence and unity.”
8–20 sec	The Walk	Photographs of families, children, groups and banners	“We walk together.” / “Inspired by Gandhi's historic march for change.”
20–34 sec	Two decades of giving	Walkathon photos, charity presentations and beneficiary photos	“20 Years of Walking for Our Community” / “Supporting local charities since 2007.”
34–48 sec	Food	Dosas, food stands, serving tables, volunteers and crowds	“A Taste of India—with proceeds supporting charity.”
48–61 sec	Culture	Children dancing, adult performances, colorful costumes and audiences	“Celebrating the richness of Indian culture.”
61–75 sec	This year's event	CASA logo/photos, walkers and the event flyer	“2026 Beneficiary: CASA for Kids”, followed by Sept. 19 • 3:30–6:00 PM • HRCC Temple

I agreed with the approach and asked it to create the video.

It created a draft video—without an audio track.

I watched the MP4 it created using the various pictures. Some images were cropped, with people's heads cut off.

I complained. It fixed them.

One image was sideways.

I complained. It fixed it.

I was concerned that the pictures made the event look as though it was aimed primarily at the Indian community, when the Walkathon is really intended for the broader Morgantown and West Virginia community.

I complained again.

It went back through the uploaded photographs, selected a more diverse set of community images, and changed some of the wording.

Now the video looked pretty good.

The Audio Track

Then I asked it to create the audio track as well, using a synthetic voice.

It did.

Now I had the entire 75-second video — well, it was 80 seconds long. The synthetic narration was terrible.

ChatGPT suggested that I record the narration myself on my phone and upload it. It gave me detailed instructions on what to say, how far to hold the phone from my mouth, and even where to pause.

It doesn't take very long to record 80 seconds of audio.

You guessed it—about 80 seconds.

I uploaded my recording, and ChatGPT spliced it into the video.

It sounded much better.

Until I actually listened carefully while watching what was on the screen.

The narration and the images didn't quite match.

So I complained again.

By now I was getting very adept at complaining.

It adjusted the timing of the images and inserted pauses in the narration so the story and the visuals lined up.

I shared that version with the other committee members. There was one suggestion about the wording. Instead of referring only to the Morgantown community, we wanted to emphasize the broader West Virginia community as well.

I re-recorded just that one sentence and uploaded it.

ChatGPT replaced the original sentence.

But you could clearly hear where the new recording had been spliced in. It sounded glitchy.

You guessed it.

I complained.

This time it added a little silence around the splice and smoothed the transition.

The result looked—and sounded—quite reasonable.

If you listen very carefully, you may still be able to tell where the splice occurred. But by then I was done complaining.

The final result is shared below.

20th Annual Gandhi Day Walkathon & India Festival | Morgantown, WV

Concluding Remarks

So that was it.

About two hours to create a 80-second marketing video—something I had never done before.

I thought that was a pretty good investment of both time and money: essentially my ChatGPT subscription and a couple of hours of experimentation.

More interesting to me was the process. I did not have to know video-editing software. I did not have to manually sort through hundreds of photographs. I did not have to figure out how to synchronize narration, images, text and music.

I mostly had to know what story I wanted to tell, look at what AI produced, and keep saying, “No, that's not quite right. Try this instead.”

Apparently, complaining is becoming an important AI skill.

And if you live within 50 miles of Morgantown, please do come to our 20th Annual Gandhi Day Walkathon & India Festival on September 19th.

Acknowledgement

The prose in this blog is mine, except where explicitly identified as ChatGPT-generated. The image was suggested and generated by ChatGPT.