When an AI-generated advertisement for David Beckham racks up over 230 million views, the marketing world stops to take notes. We are no longer in the era of 'experimental' AI video; we are in the era of high-fidelity, high-conversion production. The barrier between a bedroom creator and a top-tier agency has collapsed, provided you have the right workflow. By leveraging tools like Sora 2 and Google's Veo3, brands are now producing Super Bowl-quality content in days rather than months. This guide breaks down the exact 'sauce' used by industry leaders like PJ Ace to turn absurd historical concepts into viral marketing machines.
The Big Idea: Leveraging Public Domain IP
The first rule of viral AI video ads is that you cannot waste time introducing your characters. To capture immediate attention on platforms like TikTok Ads Manager, you need recognizable intellectual property. However, using modern copyrighted characters is a legal minefield. The solution? Public domain IP and historical icons.
By using figures like Marie Antoinette, settings like the Titanic, or events like the eruption of Pompeii, you tap into a universal cultural shorthand. Everyone knows the stakes of the Titanic; you don't need a five-second setup. You can jump straight to the 'absurdity'—like a passenger trying to sell life insurance on a sinking ship. This juxtaposition creates a 'lean-forward' moment for the viewer, making them 80% more likely to stop scrolling than they would for a generic corporate greeting, according to recent social media engagement studies.
"The key to virality isn't just the AI tech—it's the comedic contrast of placing familiar historical figures in ridiculous modern financial situations."
Scripting for Absurdity: Iterating with ChatGPT
While AI can generate video, it is notoriously bad at being funny on its own. To build a script that converts, you must use ChatGPT as a brainstorming partner rather than a lead writer. The workflow starts with a 'loose concept' phase where you ask the AI for iconic historical moments involving bad advice.
For example, a script for a financial app might look like this:
- Scene 1: A real estate agent in Pompeii selling a 'hot' new time-share next to Vesuvius.
- Scene 2: A mother telling her child their college fund is entirely invested in Beanie Babies.
- Scene 3: A passenger on the Titanic celebrating that he got 'two boat rides for the price of one.'
The goal is to find one great line out of 50. You iterate on these prompts until you find the 'funny.' Remember: AI is a tool for volume-based creativity. You aren't looking for a finished script; you're looking for the spark that allows a human director to refine the comedic timing. Once the script is locked, you can move it into a collaborative environment like Notion to finalize the shot list.
The Shot List Workflow: Organizing in Figma and Rev
The biggest mistake creators make is going 'raw dog' from text to video. This leads to morphing faces and inconsistent characters. To achieve cinematic quality, you must follow an Image-to-Video (I2V) workflow. This involves generating high-fidelity still frames first, then animating them.
Start by organizing your project in Figma. Create a board for every scene. For each shot, use a tool like Rev (app.rev.com) or Nano Banana. These platforms are superior for character consistency because they offer iterative prompting. You can select an image of a 'Roman real estate agent' and ask the AI to 'show him in a close-up' or 'change his expression to nervous.' This allows you to build a visual storyboard with 100% character continuity before a single second of video is rendered.
| Workflow Step | Tool Recommended | Why it Matters |
|---|---|---|
| Scripting | ChatGPT | High-volume ideation and historical research. |
| Visual Boarding | Figma | Keeps the entire team aligned on the visual 'vibe.' |
| Image Generation | Rev / Flux | Superior photo-realism and character consistency. |
| Animation | Veo3 / Sora 2 | Handles complex motion and character dialogue. |
Animation Best Practices: Veo3 vs. Sora 2
Once you have your 'master frames' from Rev or Figma, it’s time to animate. This is where the choice of model becomes technical. Currently, Veo3 is the gold standard for lip-syncing and character performance. If your ad relies on a character speaking a punchline, Veo3’s 'fast mode' provides 80% of the quality at a fraction of the cost, ensuring the mouth movements don't look like an old dubbed movie.
However, for scale and physics-defying shots—like a volcano erupting or the Titanic breaking—Sora 2 is the powerhouse. Sora's ability to understand complex world physics means objects move naturally. A pro-tip from PJ Ace: if the AI doesn't know what an object (like a Beanie Baby) looks like, use a 'picture-in-picture' hack. Upload a small reference image of the object alongside your character frame; the AI will use that reference to 'outpaint' the scene correctly.
For brands sourcing talent to execute these complex animations, platforms like Stormy AI can help you discover and vet UGC creators who specialize in AI-assisted video production, ensuring your campaign is handled by experts who know these technical nuances.
"We've moved from 'text-to-video' to 'intent-to-video.' The models now understand the physics of a scene better than most amateur directors."
Editing and Sound: Assembly in CapCut and Premiere
The final 10% of the work provides 90% of the 'realism.' Once you have your 5-10 second clips from Veo3 or Sora, you need to stitch them together. For most social-first ads, CapCut is the fastest tool for adding captions and quick cuts. If you require more advanced color grading or AI-enhanced skin detailing, move the project into Adobe Premiere.
Sound is the 'secret ingredient' for AI video. AI-generated video often feels 'dead' because it lacks environmental noise. While Veo3 can generate some sound effects, you should use Epidemic Sound for a consistent music bed. Do not let the AI generate your music shot-by-shot, or the tone will shift every three seconds. Use the AI for foley (the sound of the volcano, the rustle of paper) but keep the music and dialogue tracks human-led or high-quality synthesized for maximum brand cohesion.
The Future of AI-Native Agencies
We are entering a period of extreme price compression in the agency world. What used to take a 10-person crew and a $100k budget can now be done by a single 'AI Cinematographer' for a few thousand dollars. This is an massive opportunity for founders to run high-volume creative testing. Instead of one 'hero' ad per quarter, you can release one high-quality AI ad per week to see what resonates with your audience.
To manage this volume, modern growth stacks are evolving. You might pair Stormy AI for creator sourcing and CRM management with Klaviyo for email automation to distribute your new video content. The companies that win in the next six months won't just be the ones with the best AI—they will be the ones with the highest volume of great ideas.
Conclusion: Getting Your Hands Dirty
The 'sauce' is out. Viral AI video ads aren't a result of a secret button; they are the result of a disciplined workflow that prioritizes historical IP, Image-to-Video consistency, and expert post-production. The technology is moving so fast that what works today in Veo3 might be automated by Sora 2 tomorrow. Your only competitive advantage is creative taste and speed of execution.
Start by picking a public domain character, drafting an absurd script in ChatGPT, and generating your first five frames using Midjourney or Flux. The keys to the kingdom are available for the price of a few subscriptions and the willingness to 'get your hands dirty' in the prompt box. Don't wait for the tools to get perfect—start creating while the novelty still drives massive organic reach.