I made an AI ad film instead of buying a pair of jeans

Social media is seeing an explosion of AI ad films across the web. Many new companies are jumping into this space – so how could I be left behind? As a screenwriter who wants to make a feature film one day, the space is promising and concerning at the same time.

To really understand how much effort is required to make an AI film, I decided to make one. My conclusion – while the video generation has itself seen a lot of improvement, putting together an engaging, cohesive video still needs a lot of effort.

The explosion of AI-generated videos

A couple of months ago, PJ Ace posted on Twitter that he made an ad film for $500 that would otherwise have taken him half a million dollars.

I used to shoot $500k pharmaceutical commercials.

I made this for $500 in Veo 3 credits in less than a day.

What’s the argument for spending $500K now?

(Steal my prompt below ) pic.twitter.com/4UH43EXDux

— PJ Ace (@PJaccetturo) May 22, 2025

The video blew up on Twitter, and it was a turning point for people like me – the ones who want to make something but are unable to do so because of a lack of skill in domains like video / audio / editing and so on.

The reason it had such an impact on me was that it was… good. It did the job, without any glitches!

I was intent on creating something using AI, with my absolute zero audio and video editing skills. I am good at writing (I think I can claim that), so decided to give it a shot!

Before we get into the process, here’s the ad I made –

I’m sure you caught a few glitches in the video. But what if I told you that generating this ad would have taken me months of convincing people to work with me, lakhs of rupees in budget and then, no ide a whether this would work?

When I saw the completed work after I was done, I had an “Oh my god!” moment.

The process of making an AI ad film

The first step was to get access to veo3. After months of saying “Gemini is bad – it will never catch up to chatGPT”, I did end up paying for a Gemini subscription.

Google Pro costs 23 Euros in Germany, so buying this subscription was a no-brainer. This is less than the cost of a pair of jeans, which I was looking to buy around the same time.

Fun fact – since I bought it, I’ve been using Gemini a lot more (and even replacing chatGPT with it for many use cases).

Step 1: Generate the AI ad film video

With that out of the way, I began. I had a very brief idea of what I needed to develop – an ad for dog headphones during Diwali – as dogs are scared of the crackers.

I wrote down a very short prompt to Gemini, telling it to detail out a video script within its constraint of 8 seconds.

Gemini created a prompt and I edited it lightly to get to this prompt:

Please make an ad for a quick commerce product based on this description:Visuals:- 0-3 seconds: Tight close-up of a dog's anxious, trembling face (e.g., Golden Retriever, Labrador). Its ears are flattened or perked nervously, eyes wide with apprehension. The background shows subtle elements of a festive, softly lit Indian home during Diwali (e.g., a blurred string of fairy lights, a hint of a rangoli).- 3-6 seconds: A hand (implied owner's) swiftly and gently places stylish, comfortable noise-canceling headphones on the dog's head. The dog's expression undergoes an instantaneous, visible transformation to one of serenity and calm – eyes softening, head relaxing, perhaps a gentle sigh. The camera holds on this peaceful expression for a moment.- 6-8 seconds: Quick, dynamic shot. A quick commerce delivery rider, wearing a bright red, branded shirt (similar to a Zomato rider's uniform) and a matching helmet, is at a brightly lit doorstep. They are rapidly handing over a small, compact, branded package (implying the headphones) to an unseen customer's outstretched hand. The focus is on the speed and efficiency of the exchange.Audio:- 0-3 seconds: Faint, muffled firecracker sounds, subtly intensifying to convey the pet's distress and the chaotic environment.- 3-6 seconds: The firecracker sounds abruptly cut off as the headphones are placed. This is immediately followed by a soft, soothing sound effect (like a gentle, calming chime or a deep, relaxed breath). A brief, peaceful musical note or melody begins to play, creating a sense of relief.- 6-8 seconds: A subtle, quick sound of a delivery vehicle arriving and departing (e.g., a scooter pulling up and then driving away). This is accompanied by a clear, confident voiceover: "Tyohaar hai! Poore parivaar ke saath manaiye." The gentle background music from the previous section continues and fades out with the voiceover.

With this prompt, I put veo3 into action. It failed the first 2 instances: the first time there was an issue with the dog’s smile, and the second time it added audio, but the audio was too bad.

Also, after the first scene, where the dog is wearing the headphone, the second scene is of a delivery boy delivering a parcel. Veo3 had a hard time generating this video well, but it got there.

I tried a few more times with minor tweaks to the prompt, but it didn’t work.

Frustrated, I decided to take things in my hand, and this is where the fun began.

Generating the video was the easy part, putting the finishing touches on it was the much harder, longer part.

Step Two: Edit the video

As someone who has only used Premiere Pro more than 10 years ago for only piecing a video together, I was scared to even approach this space – almost ready to give up.

After some googling, I found out DaVinci Resolve as a free video editing software. Intimidated, I still went ahead and downloaded it. The tool was complicated to begin with. but like they say – all you need in life is to know how to find the right video on YouTube.

You know what the hardest part about stitching the video together was? Adding the text you see on the screen towards the end. It took me almost 3 hours to figure it out – and I did a pathetic job. If you are an editor reading this and want to help fix this, get in touch.

Step Three: Voiceover and audio

My voice is very soft, and not “ad-like” at all. I was worried it may end up turning people away from the product instead of selling it.

AI to the rescue!

I learnt about ElevenLabs and decided to give it a spin. This felt like the easiest part of the whole process. All I had to do was to write what I wanted to narrate and select a voice on ElevenLabs. I started by writing the text and using ElevenLabs’ text-to-speech, but the voice was seeming robotic.

Instead, I recorded the pitch in my own voice and asked ElevenLabs to modify the voice to the one you hear in the ad.

The last part of the puzzle was the background music. A friend recommended that I check out Royalty Free Music on YouTube Sound Library. After browsing through for about an hour, I was able to find the tracks I needed.

Putting them together in Davinci Resolve took another hour or so, before I was ready to export the video!

Rediscovering video editing was a big part of this journey. The inertia I had about video editing was somewhat alleviated by the amount of content available on YouTube for Davinci Resolve.

TradeoffsThe 8 second video limit of Veo had a major constraint in terms of what I could do. That said, many people are using this constraint to their advantage and I need to find more creative applications of this.The background music quality is not very good at the moment. It’s not clear what happened in the video until you see it twice. HUGE scope for improvement here.My original prompt was for the dog to smile towards the end of the video – but veo kept missing on this detail again and again, until I gave up. Concerns with AI generated films

One of the most immediate reaction I heard when I showed the film to a friend was “You are eating the jobs of photographers and all the artists.”

Since then, I’ve given it a thought and have concluded that the eventuality is that AI is going to be opening a lot more creative doors for people who weren’t previously able to create (like myself). I don’t mean to disrespect any of the artists. I certainly don’t think what I’ve made is good quality. At best, it is good for showing off among friends and learning.

I think the space is rapidly evolving and we are not yet seeing the ill-effects, so I don’t want to jump to a conclusion that “this will change filmmaking forever”, neither do I think that we shouldn’t use this tech.

A year down the line – we’ll end up somewhere in the middle – where we’ve figured out how to use the best of both worlds – use AI where it’s good, and use humans where they are: in being creative and telling emotional stories.

What’s next?

With this “experiment” concluded, I am looking for broader applications of AI video generation for my personal projects & for commercial uses too. Haven’t spent too much time in this space after I made this video, but I wish to get back to this soon.

Right now, my focus is on making some automations using n8n for my own personal use cases. Planning to write about them a lot more! Stay tuned to learn more about my experiments.

I am a novice at almost everything, so I’ll try my best to not intimidate you.

Subscribe

If you have any questions, shoot me an email at hemantjoshiwrites@gmail.com

The post I made an AI ad film instead of buying a pair of jeans appeared first on Hemant R Joshi.