The production cost of the AI-generated ad broadcast during the NBA Finals was approximately 290,000 yen, a 95% reduction from conventional video advertising.

A video ad generated using AI was shown during Game 3 of the NBA Finals, held on June 11, 2025 local time. This video, which had a huge impact on NBA Finals viewers, cost only $2,000 (about 288,000 yen) to produce, a 95% reduction in costs compared to traditional video ads, and it was revealed that it only took two days to produce.
An AI-generated ad aired during NBA finals and cost just $2,000 | Mashable
The video ad in question is below. The ad was released by 'Kalshi,' a service that predicts the outcome of all kinds of events, including financial markets, sports game results, and election results, and then trades those predictions.
The world's gone mad pic.twitter.com/VmoDLbwk1v
— Kalshi (@Kalshi) June 11, 2025
The Disney approved our insane AI ad to run during the NBA Finals 🤣 - YouTube
It's clear from watching the video that this was made using generative AI. The creator of this video, YouTuber and videographer PJ Ace , revealed how he created this video on his X account. Ace explained, 'Kalshi hired me to make the most insane NBA Finals commercial possible, and network television approved this GTA-style crazy ad. High-dopamine Veo 3 videos will be the advertising trend of 2025. I'll show you how I made this video in just two days with prompts.'
Kalshi hired me to make the most unhinged NBA Finals commercial possible.
— PJ Ace (@PJaccetturo) June 11, 2025
Network TV actually approved this GTA-style madness 🤣
High-dopamine Veo 3 videos will be the ad trend of 2025.
Here's how I made it in just TWO DAYS 👇🏼 (Prompt included) pic.twitter.com/XcT3m7CROL
Ace's video generation process is as follows: Ace wrote, 'I used this workflow to create a video that was played more than 30 million times in three weeks,' but it is unclear which video this 'video that was played more than 30 million times in three weeks' is.
1: Write a rough script
2: Use Gemini to create shot lists and prompts
3: Paste into Veo 3 ( Flow )
4. Edit with video editing software such as Capcut, FCPX, or Premiere
My Veo 3 viral video process is very simple.
— PJ Ace (@PJaccetturo) June 11, 2025
I've generated 30M+ views in 3 weeks using this exact workflow: pic.twitter.com/l8BGyLa9Bp
'Kalshi asked me to create a spot about people betting on different markets, including the NBA Finals. She said the best Veo 3 content is content with crazy people doing crazy things while promoting your brand. They love GTA VI. I grew up in Florida, so I guess the idea came naturally.'
Concept:
— PJ Ace (@PJaccetturo) June 11, 2025
Kalshi asked me to create a spot about people betting on various markets, including the NBA Finals.
I said the best Veo 3 content is crazy people doing crazy things while showcasing your brand.
They love GTA VI. I grew up in Florida.
This idea wrote itself 😂 pic.twitter.com/9BwrVssI6s
Below is the 'rough script' that Ace created. It describes some of the lines that the characters will say in the video. Ace explained that he did not write this script by himself, but 'co-wrote it with Gemini and ChatGPT.' According to Ace, the chat AI helped him come up with ideas and simplify the script.
Script:
— PJ Ace (@PJaccetturo) June 11, 2025
I asked their team for a few dialogue bits they wanted to include.
Then I came up with 10 wild characters in unhinged situations to say them.
I co-write with Gemini (or ChatGPT) asking it for ideas, picking the best ones, and shaping them into a simple script. pic.twitter.com/GZhtamjqhZ
Next, Gemini would input the script, which would create prompts to create the footage needed, which would then be fed into the Veo 3 to create the video. 'I usually set it to return five prompts at a time, as any more than that caused the quality to drop,' Ace writes. 'Each prompt had to completely describe the scene, as if the Veo 3 had no context for the shots before or after it. To ensure consistency, I had to explain the setting, characters, and tone each time.'
Prompting:
— PJ Ace (@PJaccetturo) June 11, 2025
I then ask Gemini to take the script and convert every shot into a detailed Veo 3 prompt. I always tell it to return 5 prompts at a time—any more than that and the quality starts to slip.
Each prompt should fully describe the scene as if Veo 3 has no context of the…
Here's an example of a prompt that Ace wrote to generate a video scene:
A raw, handheld, medium-wide shot of Miami's crowded Strip at night, shot like raw street footage. A white man in his late 60s strides confidently down the sidewalk, surrounded by tourists and club-goers. He's smiling broadly, his belly proudly protruding from his cropped pink T-shirt. He's wearing extremely short neon green shorts, white tube socks, well-worn sneakers, and a giant, sequin-encrusted foam cowboy hat. His leathery, tanned skin glows under the neon lights.
In one hand, he holds a trembling little Chihuahua like a precious accessory. As he struts along, he turns slightly toward the camera and exclaims with confidence and joy, 'Indiana has that dog!' Close behind him are two older women dressed in 1980s-style attire, both in sparkly workout leotards, chunky sneakers, and huge plastic sunglasses. Their hair is still in curlers and they both wear clear plastic shower caps. One of them takes a sip from an oversized novelty-style margarita glass while the other waves at passing cars.
All around them, the Strip is bustling: people take pictures with their phones, scooters whizz by, music blares from a nearby balcony. Neon signs flash overhead, casting a blast of electric color all around. Around the trio, the crowd disperses, looking a mixture of surprised and confused.
Below is a scene from a video that appears to have been created using the prompt above.
Ace's tips for generating videos with Veo 3 include 'running five prompts at a time in fast mode ($0.20 each)' and 'if the generated video isn't right, paste the prompt into Gemini, ask them to change it, and generate the video again in Flow.' He also recommended writing 'yell out loud' to make a character shout out, or writing the lines in capital letters.
Veo 3 / Flow tip:
— PJ Ace (@PJaccetturo) June 11, 2025
Run 5 prompts at a time in fast mode ($0.20 each).
If a shot isn't right, paste the prompt into Gemini, ask for changes, then try again in Flow.
Leapfrog ahead while waiting and then repeat until the shot list is done. pic.twitter.com/DVZPAcbYXe
'It took about 300-400 production runs to create 15 usable video assets. It took one person 2-3 days to do that. That's a 95% cost reduction compared to traditional advertising.'
Cost and Time:
— PJ Ace (@PJaccetturo) June 11, 2025
This took about 300–400 generations to get 15 usable clips. One person, 2-3 days.
That's a 95% cost reduction vs traditional ads.
Speaking about the future of video advertising, Ace explained, 'Just because it's cheap doesn't mean that anyone can create video ads using AI. I've been a director for over 15 years, and brands still pay a premium for taste. In the future, we'll see small teams creating viral content related to a brand every week, achieving 80-90% of the results at a much lower cost.'
The Future of Ads:
— PJ Ace (@PJaccetturo) June 11, 2025
Just because this was cheap doesn't mean anyone can do it.
I've been a director 15+ years. Brands still pay a premium for taste.
The future is small teams making viral, brand-adjacent content weekly, getting 80 to 90 percent of the results for way less.
Ace writes about the skills that video ad creators of the future will need: 'Right now, the most valuable skill in the entertainment and advertising industry is comedy scriptwriting. If you can make people laugh, they'll watch the whole ad, they'll be interested, and some of them will become customers.'
What's the Moat for Filmmakers?
— PJ Ace (@PJaccetturo) June 11, 2025
It's attention.
Right now the most valuable skill in entertainment and advertising is comedy writing.
If you can make people laugh, they'll watch the full ad, engage with it, and some of them will become customers.
in Video, Posted by logu_ii