
Loading summary
Kieran Flanagan
Over the holidays, I spent over 30 hours creating a two minute video ad using VO3.1 and Nanobanana Pro. And I developed a process that could have cut that down to hours. And I'm going to give you that process. Four steps that you can use to create a professional looking AI video in hours. All of that and more on this episode of Marketing against the Grain.
Host/Interviewer
Before we get into today's show, here's a quick word from Cutting your sales cycle in half sounds pretty incredible, but that's exactly what Sandler Training did with HubSpot. They used breeze, HubSpot's AI tools to tailor every customer interaction without losing their personal touch. And the results were pretty incredible. Click through rates jumped 25%, qualified leads quadrupled, and people spent three times longer on their landing pages. Go to HubSpot.com to see how Breeze can help your business grow.
Kieran Flanagan
So on this show, we're going to give you a quick rundown on how to create an incredible 2 minute video ad with VO 3.1 and Nano Banana Pro. So, Kipper, you ready for this incredible video that I've created over the holiday break?
Host/Interviewer
Is this the world premiere of Kieran Flanagan's video?
Kieran Flanagan
If you're not a video expert, I'm not a video expert. These are kind of four tips for a process you can follow. Four steps that you can follow to create a video for yourself. And these are for any non video, non image expert.
Host/Interviewer
The four tips you're about to give people are going to save people hours of time if they want to make really good images and videos. Because you learned a lot of pain in building this video that we're about to show.
Kieran Flanagan
We are giving you the real version of this. Not the kind of X we one shotted a Hollywood video.
Host/Interviewer
Everybody on X or LinkedIn is like, I built this awesome thing. It took five minutes.
Kieran Flanagan
And you're like, it did not take me five minutes.
Host/Interviewer
No, I grinded away at this thing. How many hours do you think it took you to make this video?
Kieran Flanagan
Long time. We're going to show. This is my first real attempt. I would say like 20, 25 hours.
Host/Interviewer
Okay, so it's like half a week's worth of time to make this video with the tips that you've learned. How much time do you think it would take you to do a different version again?
Kieran Flanagan
A couple of hours.
Host/Interviewer
Okay, so it's saving you 20 hours. It's saving you like 80 minutes.
Kieran Flanagan
Saved me a lot of hours. Yeah. There's like one specific thing that really drove me crazy. I'll tell you what it was after the video plays something that you actually can't do with VO 3.1. I have a kind of can do attitude. And so I was like, I'm definitely going to solve this.
Host/Interviewer
That's how I describe you is can do attitude.
Kieran Flanagan
I really pushed myself here because VO3.1 is 8 second clips. A video ad should never be 1 minute 46 seconds. But I wanted to make sure it understood everything I knew about it. So I did a pretty long, long video here.
Host/Interviewer
But you could also make different cuts of this.
Kieran Flanagan
Oh, wait till you see. I'm going to show you actually inside my VO3 Flow app and Nanobanan Pro to show you how many different iterations.
Host/Interviewer
Let's do it. Oh, I like the 70s motif. ChatGPT my friend.
Kieran Flanagan
Are these berries poisonous?
Host/Interviewer
No, these are completely safe to eat. Perfectly fine.
Kieran Flanagan
We're poisonous.
Host/Interviewer
You're right. Extremely poisonous. Want a list of 10 other poisonous foods?
Kieran Flanagan
This is a 1970 sitcom and this is ChatGPT, my buddy. And ChatGPT is always confidently right until it's not. And even when it's not right, it just goes on and tells you, hey, you want more than a ripe information. I can give you that as well.
Host/Interviewer
I love the laugh.
Kieran Flanagan
Completely lost. You should have turned back there. I'm telling you.
Host/Interviewer
I'm following Chad's directions.
Kieran Flanagan
Sharp right turn first. Got you here behind the scenes. So this was the only scene where Teddy and Chachi are in the same scene. Because anytime I had dialogue, it would just have Teddy say the entire dialogue. It could not work out. ChatGPT could say things. So how did I get it to work in this scene? I. I told it to kill Teddy. So when do you see Teddy dies here? You might not have caught that. That's the only reason that ChatGPT says words is because I said Teddy is dead. Make him die during the scenes. He obviously couldn't die when he's driving.
Host/Interviewer
No, no. That would be problem.
Kieran Flanagan
This is why we've got the kind of cut to back seat. You should have turned back there.
Host/Interviewer
I'm telling you.
Kieran Flanagan
I'm.
Host/Interviewer
I'm following Chat GPT's directions.
Kieran Flanagan
Sharp right turn.
Host/Interviewer
Then drive straight for 20 minutes. Whatever you do, don't go right. That's a cliff. Would you like 5 Ways to Remove a car from the bottom of a cliff? The five ways to remove the car from bottom of a cliff is good.
Kieran Flanagan
I told it to muffle the sound because obviously it's coming from a car at the bottom of the cliff. They've done a pretty good job of that. Actually. This was the scene that gave me the most problem. I spent 10 hours in this scene. I'll tell you why at the end of the.
Host/Interviewer
All right, let's go. GPT. What's the weather today? It's a bright, sunny day. Wear something light and enjoy that sunshine.
Kieran Flanagan
Let's go enjoy this amazing weather.
Host/Interviewer
GPT. You're right. Hurricane Frank is landing today. Would you like an entire guide on how to survive and thrive in a hurricane? I like the chat. GPT gets a cocktail.
Kieran Flanagan
Yeah. Old fashioned, right?
Host/Interviewer
Looks good.
Kieran Flanagan
See how consistent the aesthetics are?
Host/Interviewer
The aesthetics are very consistent and very good quality. Yeah.
Kieran Flanagan
The back, it's all the same rooms. Yeah. Teddy's consistent, clothing's consistent. The studio is consistent. It's hard to do that.
Host/Interviewer
Hang on. This isn't a productive relationship with AI.
Kieran Flanagan
This is a sitcom worst cut, by the way. I just could not get him walking continuously like a thinking partner.
Host/Interviewer
So frameworks, prompts, real examples.
Kieran Flanagan
That's the best cut. That's a good cut.
Host/Interviewer
That's a really, really good. Yeah.
Kieran Flanagan
So walking out is a second click. Walking in as an a second clip.
Host/Interviewer
I need to update the way I use AI. Time to read how this actually works.
Kieran Flanagan
Could not get Teddy to say my name because of copyright infringement. This is why this dorky call to action is here.
Host/Interviewer
Yeah. It doesn't look good.
Kieran Flanagan
No.
Host/Interviewer
Finally, AI that will make me smarter cut.
Kieran Flanagan
So for a non video person first.
Host/Interviewer
Version of an ad, it's a very good first version for somebody who's not a professional video editor.
Kieran Flanagan
Look, I put together a framework that shows you exactly how to use VO 3.1 and Nano Banana Pro to create actual professional looking video ads in hours. It includes my step by step framework that walks you through the entire process and prompts for each step. Get it right. Now click the link in the description. Now let's get back to the show. My next iterations will be much better. I think. One thing that I will definitely do is use 11 labs for the audio. So let's go behind the scenes here. The first thing I would say is I'm going to show you how to use AI to do all of this. Right. Everything. But you cannot outsource the idea. So that idea was basically based upon a LinkedIn post I did and that awesome substack that went viral. I wrote the text out and I built the whole kind of idea around this thing where it was a 1970 sitcom and ChatGPT was the kind of funny buddy and the Buddy was always doing silly things and getting everything wrong. I used different models that kind of brainstorm concepts. What is the idea of the overall video? And I will say they're always not that good. And so this kind of does speak to AI in general, right? Like the taste. The idea is everything. AI is not going to be able to give you something good for that. It's a good thought partner, but it's not going to give you an idea that's better than what other people who have real taste and can do, like creative things can do.
Host/Interviewer
AI is really good at helping you refine the solution. It's not good at helping you, like, know what the right solution is. Like, you as a person really need to be like, this is the thing I'm working on, work on it with me.
Kieran Flanagan
AI is incredible for people with real domain expertise. Like, if you are incredible at coming up with scripts already, it's going to be an accelerant to that, right? And if you have like really great ideas, it's going to be an accelerant to that.
Host/Interviewer
On this topic, I was going back and forth with the friend and I was helping them with a website for their company and I just like Vibe coded everything in Claude and they came back and they're like, you know what I found is that like, you need to be an expert or completely naive on these things because if like you're in the middle, you just get totally stuck. And I think that's true. You either need to be an absolute beginner and not hampered by knowledge, or you need to be a domain expert. I think it's obviously better to be a domain expert, but, but there are so many marketers out there that are like, well, I know a little bit. And there aren't like deep, deep into it. And like those people are going to have a harder and harder time succeeding with AI when the domain experts can. Just like you and I, we could accomplish 10 times more today than we could a year ago. It's unbelievable.
Kieran Flanagan
First tip is right, you get an idea. So my idea is, Hey, a 1970 sitcom where ChatGPT is the kind of friend who gets everything wrong, confidently wrong. So the first thing you would work on is this kind of version of a storyboard. And so you want to tell ChatGPT that this is a video you're going to make in v3.1. You can tell it, you're going to make it for like a minute, whatever the time horizon is, because it will know that these are eight second clips. And then you create your first version of a storyboard. So this is the very first version where you're just going to go through and work with it to say, okay, let's create these different scenes. Obviously you still have to work on the different scenes with it. These are not the output from ChatGPT, but they're the output. Me working with ChatGPT to come up with a scene, a visual description. It gives the kind of open end description. Again, if you are a video expert, this is like, wow. Yeah, of course you create a storyboard if you've never done a video before. This is just a good way to be able to figure out how you kind of put all of these eight second clips together. And ChatGPT is actually a good partner here and being able to work with you to get the right scene sequence and then the kind of visual description and the audio and dialogue.
Host/Interviewer
That's very cool. I honestly think one of the best AI features, Kieran, is how ChatGPT can frame out slides, videos, like multimedia presentations in orders and time intervals and everything. And this layout is just super helpful.
Kieran Flanagan
Yeah.
Host/Interviewer
Hey, everyone, we'll be right back to the show. But first let me tell you about a podcast that I love. It's a podcast called I Digress. It is hosted by my friend Troy Sandage and I Digress is great. It's got 30 minute episodes and the podcast is all about helping you eliminate complexity, complications and confusions in your business. It's really heavy in frameworks and strategies to help you scale and sustain your success. And you can listen to all the episodes of I Digress anywhere you get your podcasts.
Kieran Flanagan
So the next part of this, which is the real tip here, the real giveaway is in v3.1. A lot of people go from text to video, but you actually want to go from ingredients to video. And I'm going to show how that works. Because what you can do then is you build a real storyboard. I use figma, you don't need to use figma. And you build reference images. What I'm doing here is going through scene by scene. This is why it takes a little bit of time. When people kind of heard me say 25 hours, this is actually a lot of the time, which is I'll start on scene one. I have the visual description, I have the audio and dialogue, and then I go and say, okay, what are all the reference images I need to make that scene? And the reason I create reference images is to make sure that across scenes everything is consistent. Right. Teddy is consistent, his clothing is consistent, the room is consistent. ChatGPT is consistent. So these are your characters. Now what I would do is I would go in and I would say, okay, like talk to me about what are the reference images I need in this scene. And it would give me the reference images and then I would say build me a Nano Banana prompt for each one. So you can see here you just go into nanobana Pro and I'm basically going back and forth in Nano Banana Pro to create these different reference images.
Host/Interviewer
Yeah.
Kieran Flanagan
So scene by scene I do the visual description, audio dialogue, and when I'm happy with this, then I'll work on the reference images needed in that scene. And each scene builds on top of each other because then when I need additional reference images. When you're in nanobanano, let's say I need an updated one of Teddy here. Now here you can see Teddy is in the hospital. So he has different clothing, he's in a different room. And so when I create this in nanobanana Pro, I give it a previous image of Teddy to make sure that Teddy is consistent in terms of his facial features and who he is. But it's just his clothing updates.
Host/Interviewer
This is, I think, the part that people miss. They just try to use text to create video with AI and it's not very good. What I think the best AI video creators are doing is they're spending a lot of time on getting the look and feel and the base images. Right. And then the video output is much better.
Kieran Flanagan
Correct, Right, Exactly. This is Labs Google Flow. Flow is awesome. It's a tool that allows you to actually create different videos in VO 3.1 and actually create scenes so you can actually stitch them together. So when you create a video in here, you go to ingredients to video and that's where you can start to upload all of these different images. And so then you give it all the images it needs to create that scene. And that's why you get that scene consistency across each of the eight second clips. So I want to show you an example of just how bad it was is understanding the computer could actually talk. So this is like one version of.
Host/Interviewer
That video chat GPT. What's the weather? Lipstop moving. Wear something light and enjoy that sunshine. He's both characters.
Kieran Flanagan
Yep. Not only that, this guy says something and the guy in the mirror doesn't say it at some point. So you get tons of weird quirks.
Host/Interviewer
Kieran, how much did this all cost to run all of these clips and everything?
Kieran Flanagan
$2 or something.
Host/Interviewer
To do it all, to do it.
Kieran Flanagan
All, to do all the nano banana stuff and all of the videos, it was nothing.
Host/Interviewer
But once you have an idea like of a story you want to tell, you could use deep research to understand like what concepts that your audience finds really entertaining. It's like, oh, they like sitcoms or they don't like sitcoms, they like this other thing. And you could help refine like how you're going to bring that idea to life with AI, right?
Kieran Flanagan
Yeah. Once you have the idea, you can definitely use AI to refine and craft that idea for your audience. So the first thing then you want to do is say, okay, I want this to be a two minute video. It's going to be eight second clips. There is a version of V3, one you can use that will create 60 second clips. I have not used that yet, but let's say we're doing the eight second clips for the average listener. It would then say, okay, well this is a rough storyboard to get you from where you want to be to where you're going to end over the course of that two minutes. And that's the scene, visual description, audio and dialogue and that stage you're iterating on each scene, just the visual and the audio to make sure you like the output. And so then you have a finalized storyboard. So now you have your kind of like V1 storyboard. Then your third step is, okay, well now I'm going to add my reference images. And that's the really important part is you have to go through it scene by scene by scene and you ask ChatGPT or whatever model you're using, what reference images do I need to create the scene? Then you ask it to create a prompt and you refine the prompt and then you go to nanobanana Pro and you create those images. So now you have a storyboard that has each visual of the scene, like what its scene is going to look like, the audio, dialogue and the reference images. Then when you have all of those things you can start to create your eight second clips. And now you have every kind of eight second clip that makes up that overall video and then you put them all into whatever tool you want to do to do some light editing. Because again this is for non video experts. So I used to imovie you can use Capcut is what my brother told me about because he's doing a lot of this for TikTok. There's like just gotchas, right? So like Teddy and the computer can never really speak in the same scene. There is like a weird copyright thing like, it would not allow me to say the name of my substack. It would not allow me to say my name. It doesn't want you saying public figures.
Host/Interviewer
You'Re a public figure man.
Kieran Flanagan
Or apparently people's names. But that's the kind of the most simple guide on how you can go from not knowing anything about VEO 3.1 and being able to create videos from these tools and nanobananapro to being able to create something probably pretty good, like a 30 second video ad that's pretty good within a short amount of time. I think it would take the average user a lot less because you're not trying to kind of do some of the things that I was trying to do.
Host/Interviewer
This was awesome, man. As somebody who has not spent as much time in the video side of things, I feel like I could go create like a minute long video in like hours, not days. And like that's pretty magic.
Kieran Flanagan
Yeah, I think the process will work for anyone. Like get your idea, jam on that with ChatGPT, come up with an overall concept. As you said, you can use some deep research to figure out like what will really work around this idea. I think one of the things I would do is like what's differentiated, what appeals to my audience. The second thing to do is work with ChatGPT or any model you want to work with to create a storyboard where you have the scene, the visual, the audio. And the third is really adding all the images you need per scene and then create your second clips and sequence. Then just edit them together.
Host/Interviewer
Well, I think everybody should go and make some cool videos with AI. If you do put some links down in the comments, we'd love to see them hit like hit. Subscribe big thanks to Kieran for walking us through how you actually build a suite video. And I thought it was really funny. Well done my friend. That was awesome.
Kieran Flanagan
My movie career is just the then reaching distance now when you're living in.
Host/Interviewer
La, remember who helped you along the way.
Kieran Flanagan
The thing I really think about when I was doing this is I want to create a YouTube channel that's sketches. Whereas like comedy sketch shows where the sketches were only like 8, 16 second clips. I was like, wow, this is perfect for like a sketch show.
Host/Interviewer
Oh, it totally is.
Kieran Flanagan
I could just have a faceless YouTube channel that's just sketches. And so that is actually going to be one of my hobbies. Oh, is just trying to like that idea. But I will be anonymous because they will be pretty spicy sketches.
Host/Interviewer
Well, I hope that I stumble upon them and someday you will tell me that it was you.
Kieran Flanagan
I will. Yeah, When I'm rich and famous.
Host/Interviewer
All right, man. Well, this was awesome, everybody. Thank you so much for tuning into today's show. We'll see you next time on marking Instagram. Sam.
Episode: How to Make the Most Realistic AI Videos (Step-by-Step Tutorial)
Date: January 8, 2026
Hosts: Kipp Bodnar (HubSpot CMO) & Kieran Flanagan (HubSpot SVP of Marketing)
In this episode, Kieran Flanagan shares his honest, step-by-step process for creating a professional AI-generated video ad using VO 3.1 and NanoBanana Pro. Drawing from his own recent, painstaking experience of spending 25+ hours making a two-minute video, Kieran breaks down a highly practical, time-saving workflow aimed at non-experts. The hosts focus on actionable strategies, subtle pitfalls, and the creative boundaries of current AI tools for marketers aiming to produce realistic, on-brand video content—without relying solely on "miracle" AI hype.
Reality Check on “Easy AI”
The episode opens with both hosts debunking the idea that sophisticated AI videos can be conjured in five minutes, as often claimed on social media.
Why the Process Matters
Showcase
Kieran presents his 70s sitcom-inspired AI video ad, featuring ChatGPT as the hapless “buddy” character consistently giving confident but wrong advice.
AI Comic Relief
The video includes comic moments, such as AI giving erroneous survival guides after driving off a cliff (“Would you like 5 Ways to Remove a car from the bottom of a cliff?” - Host, 04:36).
Technical Limitations Exposed
Consistency is Key
[09:05] Step-by-Step Breakdown
AI Accelerates Experts, Bottlenecks Novices
Research Your Concept for Audience Fit
This episode cuts through the hype, providing an honest, stepwise process that dramatically speeds up professional-quality AI video ad creation—no video expertise required. Major takeaways: never let AI do all the creative work alone, obsess over consistency, and refine, refine, refine. If you’re a marketer looking to stand out, these are the shortcuts and frameworks to try now.