A lonely AI filmmaker, a state -of -the -art liberal video model, and a national TV spot during one of the biggest sports events of the year. This is not the plot of a science-fi film; This is the new reality of advertising, and it was built in just 3 days.
TLDR:
- First of its kind: An AI-birthted commercial Kalashi for events-beating platforms was broadcast Google’s generic video model, Veo3, and nationally during the NBA final.
- Rigid cost reduction: AI filmmaker PJ ACCETTURO created a 15-clip advertisement in just 3 days, resulting in an estimated 95% cost compared to traditional commercial production. It costs $ 2,000
- “Gemini-to-Veo 3” Workflow: The advertisement was manufactured using a simple but powerful 4-phase process: Rough script, AI-in-operated early generation with Gemini, video generation with VO3, and final editing in standard software such as Adobe Premier.
- The future is agile: The project indicates a change towards small, highly efficient creative teams, which take advantage of AI, which quickly and economically prepare a high-trip, brand-blessing material.
- Human skills are still key: Despite the technical jump, the manufacturer emphasizes that professional taste, director’s experience, and, the most important thing is that in the AI era, there are new mats for creative in the AI era…. And I agree!
In a historic moment for both advertising and AI, a commercial generated by AI aired on National Television this week during the NBA final. Advertisement, Kalashi, Self-known “AI filmmaker” for the event-setting market was the brainchild of PJ Esteturo, which was hired to build a place about people who bet on everything from sports to current events.
Is advertising Here,
The result is a will for generative video technology’s abilities to grow rapidly. In a detailed post on X, ACCETTURO unveiled the shocking efficient process behind the advertisement, which takes advantage of Google’s powerful new text-to-video model, Veo3. This achievement underlines the breaknake speed, which comes a few weeks after the VEO3’s public debut, on which the AI is being adopted for the commercial use of high-day.
AI-Interactive Creative Workflow: In the days of script to screen
The process of Accetturo, which claims that he claimed that it has been viewed more than 30 million times in various projects in just three weeks, a masterclass in Manav-AI collaboration. He breaks his viral video workflow in four simple stages:
- Write a thick script: The process begins with a fundamental creative idea.
- Use Gemini for signals: The Gemini of Google is used to present the script in a wide shot list and generate specific signals for video models.
- Absolutely generated with veo3: The signals to generate raw video clips are fed in the veo3 (through the Google flow).
- Edit in Capcut/Premiere: The final AI-generated clips are assembled, timely and polished using industry-standard video editing software.
Co-writing with a machine: Crafting to vision
For Kalshi spots, the creative process began by installing some major dialogue snipet. From there, ACCETTURO collaborated with Gemini, in which he invented 10 wild characters 10 wild characters “without any conditions.
“I co-write with Gemini,” Acsetturo explains, “asking it for ideas, choosing the best people, and shaping them in a simple script.” This partnership allows for rapid consideration by combining human creative direction with immense imagination of a large language model.
Prommping is new direction
There is a significant translation layer indication between the script and the screen. Accetturo has refined a specific method for this, which is using Gemini to convert each shot from the script in a high wide paragraph for veo.
“I then ask Gemini to take the script and convert every shot to a wide VO Prompt,” he notes. “I always ask it to return 5 signs at a time – even more and the quality starts slipping.”
The key, he emphasizes, is to assume each signal as a standalone instruction, provides VO with complete reference every time to maintain stability in character, settings and tones.
Here is an example of a detailed indication used for advertising:
A handheld medium-wind shot like raw street footage on crowded Miami strip at night. Surrounded by an old white person, tourists and clubgors in the late 60s, below the pavement. He is smiling from ear to ear, his stomach proudly sticks out of a cropped pink t-shirt. He wears a large-scale foam cowboy hat with very small neon green shorts, white tube socks, beet-up sneakers, and sequins on it … As soon as he walks, he turns a bit towards the camera, still shouts with mid-straight, and full confidence and joy: ‘Indiana found that dog in’ M ‘!
Tips, tricks and bottomats
Some insider tricks are required to work with Veo3. Accetturo recommends running the prompt in “fast mode” and repetitions quickly. If a generation is not correct, he suggests to paste the original prompt back to Gemini, ask for specific changes and try again. He also notes some current quirks, such as sometimes unwanted subtitles and the need to use phrases such as “shouting on top of his lungs” or all-cap dialogues to shout from AI characters.
While the Veo3 does not yet support the varnas in several shots, extremely detailed descriptions in each prompt can create a strong confusion of continuity.
The most staggering figure with this experiment is efficiency. “It took around 300–400 generations to get 15 usable clips,” Acsetturo writes. “One person, 2-3 days. It is a 95% cost reduction versus traditional advertisement.”
The future and creative “gap” of the advertisement
Accetturo is early to indicate that this revolution does not magic of the end of creative professionals. “Just because it was cheap it does not mean that anyone can do so,” they say. “I’m a director [for] 15+ years. The brands still pay a premium to taste. ,
He imagines the dominated future of small, agile teams producing “viral, brand-oriented materials weekly”, which achieves 80–90% impact for a fraction of cost.
So, what is defensive skill for filmmakers and advertisers in this new paradigm? According to ACCETTURO, it is not technical skills but creative instinct. I agree and this is a good proxy for a human+ AI future.
“The most valuable skill in entertainment and advertising is comedy writing,” He concludes. “If you can make people laugh, they will see full advertisements, connect with it, and some of them will become customers.”
Kalshi NBA Final Advertisement is more than just a clever commercial; This is a dispatch from the future of media construction – a future that is coming fast.
“AI Film Producer” PJ ACCETTURO NEWSLETER is here: https://pjace.beehiiv.com
Jean-Mark is a successful AI business executive. He enhances growth and speed for AI -operated solutions and starts a computer vision company in 2006. He is a recognized speaker at AI conferences and MBA from Stanford.
