My AI Videos Hit 1M+ Views (Veo3 + Sora 2 Demo)

Key Concepts

• Viral AI Video Ads
• AI-Native Agency Workflow
• Comedy-First Ad Strategy
• Scripting & Concept Generation
• AI Image Generation (Rev, Nano Banana, Freepick)
• AI Video Animation (V3, Cling, Luma Labs, Seed Dream/Dance, Miniax)
• Editing Software (CapCut, DaVinci Resolve, Premiere, Final Cut)
• Sora's Disruptive Impact
• IP & Likeness in AI Content
• Image-to-Video vs. Text-to-Video Workflow

---

Introduction to Viral AI Video Ad Workflow

PJ Ace, dubbed the "viral AI ad man," presents his comprehensive workflow for creating AI video ads that achieve hundreds of millions of views, enabling his AI-native agency to scale to six- and seven-figure revenues within months. The session promises a "full open kimono" approach, sharing all prompts, tools, and screen demonstrations. The core objective is to empower viewers to create viral AI videos, a skill for which people typically charge thousands of dollars.

Core Philosophy and Strategy for Viral AI Ads

PJ Ace emphasizes a specific strategy for AI video ads, particularly when working with larger companies, to mitigate potential negative pushback regarding AI's impact on jobs.

• Comedy-First Approach: Ads should be "ridiculous," "comedy first," and primarily consist of jokes, with minimal brand integration at the end. This makes the content feel like a "Super Bowl commercial."
• Mitigating AI Disruption: By making ads entertaining and clearly absurd, viewers are more likely to watch the full ad and less likely to criticize the use of AI, as the content's comedic nature makes it less threatening.
• Three Pillars for Viewer Engagement:
  1. Existing Public Domain IP: Utilize recognizable, public domain intellectual property (e.g., Pompeii, Marie Antoinette, Titanic) to leverage audience familiarity.
  2. Juxtapositions: Incorporate comedic contrasts (e.g., Beanie Babies as a college fund, a Titanic life raft offering "two boat rides for the price of one") to create "lean-forward" content that encourages sharing.
  3. Internet-Native/Trending Content: Integrate current internet trends (e.g., meme coins, NFT bros) to ensure relevance, especially on platforms like X.

Case Study: Origin Financial Ad (2 million views) This ad exemplified the strategy by showcasing "bad financial advice throughout history" in chronological order. Examples included a Pompeii time share, Marie Antoinette's cake obsession, the Titanic, Beanie Babies as a college fund, Enron and Blockbuster diversification, and meme stocks. The ad's success was attributed to its "extra" and humorous nature.

Case Study: Kelshi Video (50 million views) A similar historical framework was used, but this ad focused on inspiring "underdog moments" (e.g., Jesus rising, David vs. Goliath, Wright brothers) to convey a message of defying odds, rather than pure comedy.

The 5-Step AI Video Ad Workflow

The workflow is broken down into distinct phases, moving from concept to final edit.

1. Scripting and Concept Generation

• Initial Phase: Begin by pitching three distinct concepts to the client.
• Team Structure: While starting as an "army of one" (wearing all hats), scaling involves working with experts like "proriters," directors, and "AI cinematographers," with PJ Ace often acting as an executive creative director.
• Tool: Google Docs is preferred for collaborative scripting due to its note-making capabilities.
• Big Idea Development: Start with "recognizable IP" to create a sense of "familiar with the foreign."
• Brainstorming with ChatGPT: Use ChatGPT to generate initial ideas and potential lines (e.g., "give me ideas for iconic moments throughout history that incorporate bad advice").
  • Caveat: ChatGPT is "not funny 99 times out of 100," but it can spark a great idea or provide a usable line.
• Iterative Process: Ideas often evolve and improve during the creation process, allowing for continuous elevation of the concept.

2. Script to Shot List Generation (ChatGPT)

• Process: Once the script is finalized, upload it to ChatGPT.
• Prompt: Instruct ChatGPT to "turn this into scenes and a shot list for each scene," requesting structured prompts for images.
• Purpose: This step is crucial for generating image prompts for each shot, which is "easier and cheaper and faster" than generating full text-to-video clips blindly. It allows for shot-by-shot iteration and client approval before animation.

3. AI Image Generation and Figma Board Layout

• Primary Tool: Rev (app.rev.com): PJ Ace highly recommends Rev for its user-friendly interface, "photo realism," and ability to generate three variations per prompt.
  • Iterative Prompting: Rev offers suggestions for prompt refinement (e.g., "move the passenger closer to the camera," "add more dock workers") and can generate close-ups or variations based on existing images.
  • Advantage: This "shot by shot" approach is a "light years ahead" improvement over older text-to-video models (like V03 in its early days) that required "burning $4 a shot" blindly.
• Alternative Image Tools:
  • Nano Banana: Uses similar technology to Rev, but Rev's interface is preferred.
  • Freepick: An all-in-one platform offering various image models (Google, Nano Banana, Flux, Crad). While convenient, its API pricing can be more expensive than dedicated tools.
• Figma Board: All generated images, including alternative versions, are organized on a Figma board.
  • Role of AI Cinematographers: These specialists take the script and director's treatment to generate numerous image options, filling the boards.
  • "Expert Mode" (e.g., David Beckham ad, 230 million views): For complex, stylized ads, hundreds of shots might be generated for a single item on the shot list, with extensive reference images for tone and style. The director then makes "selects."
• Enhancer.AI: A program that can be used for a second pass to add fine details like acne or "rough up" faces, counteracting the "AI glow" for more realistic character appearances.
• Cost: Image generation is "practically free," with unlimited generation available on platforms like Freepick for around $20/month.

4. V3 Animation (Frames-to-Video)

• Primary Tool: V3 (Google's model): Considered the best model for animating characters that talk, offering "really realistic" motion and character performances.
• Process:
  1. Download selected images from Rev/Figma.
  2. Upload the image as the "first frame" into V3.
  3. Input the dialogue and an animation prompt (e.g., "move camera left to end on the ship").
  4. ChatGPT for Animation Prompts: ChatGPT can assist non-filmmakers in generating dynamic camera movements (e.g., "move camera left to end on the ship" is sufficient).
• Other Animation Tools (for non-talking characters): Cling (Kling), Luma Labs, Seed Dream/Dance (by ByteDance/TikTok), and Miniax are mentioned as similar alternatives.
• Google's Indemnification: V3 (Google) is favored for commercial client work due to its strong indemnification policies, ethically trained data, and robust training data.
• Advanced Trick: Outpainting/Picture-in-Picture: To achieve specific visual effects, a reference image (e.g., Beanie Babies with Ty tags) can be uploaded within V3, and the AI prompted to "uncover this" or "pan over to this ship," allowing the AI to "outpaint" or integrate the reference image into the animation.

5. Editing

• Process: Generated clips are sequentially placed into a timeline.
• Simplicity: AI video models like V3 often add sound effects and dialogue, significantly simplifying the editing process.
• Music: It's crucial NOT to let AI generate music for each clip, as it will create disparate tracks. Instead, use external music libraries.
  • Recommended: Epidemic Sound ($9/month for unlimited songs) and Storyblocks.
• Editing Software:
  • Free Options: CapCut (online, TikTok quality) and DaVinci Resolve.
  • Paid Options: Adobe Premiere ($19/month) and Final Cut (described as "iMovie on steroids" and "probably the easiest platform").

Case Study: RAMP Ad

• Product: Financial services credit card company (expense management).
• Concept: Portraying audit season as a horror movie, with the tagline "Audit this."
• Workflow Evolution: Initially, the team attempted text-to-video, but the results were unsatisfactory ("faces were morphe and it looked like shit"). This led to a complete re-do using the image-to-video workflow, which significantly improved the "base images" and overall "cinematic" quality.
• Process: Similar to the Origin ad, writers created a script, a director developed a shot list, and AI Director of Photographers generated variations for each shot (e.g., zombies, horrified reactions).
• Fidelity and Realism: The image-to-video approach "upped the fidelity" and made characters appear "so much more cinematic" and less "AI generated." Enhancer.AI was also used to add imperfections for greater realism.

The Future: Sora's Disruptive Impact

PJ Ace acknowledges the imminent disruption from Sora, OpenAI's text-to-video model.

• Sora's Capabilities: Sora can auto-generate scripts, images, video, sound effects, and music in 10-second clips.
• Future Developments: OpenAI is expected to extend clip lengths (30s, 60s), improve character consistency, and enable individual clip tweaking.
• Disruption to Agencies: Sora will drastically reduce production timelines from 6-8 weeks to "a week or less," posing a significant challenge to existing AI agencies.
• Agency Adaptation Strategy:
  • Volume & Pricing: Agencies will need to lower prices and increase volume, potentially shifting to retainer models with brands.
  • Role Specialization: Dependency on AI cinematographers and animators will decrease, while the focus will shift to "comedy writers and supervising directors."
• Opportunity for Brands: Brands will benefit from being able to release new ads weekly at lower price points.
• The Limiting Factor: Great Ideas: In an age of automated content generation, the ultimate differentiator will be "great ideas" and "scroll-stopping ideas that aren't thinking like everyone else."
• Meme Branding and IP:
  • Stephen Hawking Example: The viral Stephen Hawking clip demonstrates the potential for "meme branding" with photoreal physics.
  • Likeness and Royalties: PJ Ace speculates that if estates of public figures (e.g., Hawking, Tupac, Kobe) opt-in and charge royalties for their likeness, it could unlock new branding opportunities.
  • Open-Source IP: Brands can leverage historical figures (Einstein, Plato) from open-source IP.
  • IP Undervaluation: PJ Ace believes IP is currently undervalued, anticipating that OpenAI and similar companies will eventually strike deals with IP holders.
  • Production Company Opportunity: Small creators and agencies can act as intermediaries for revitalizing old IPs (e.g., Thundercats, He-Man) that lack internal expertise in AI prompting.
  • Cost Compression: Producing new episodes for old IPs could range from $30,000 (low end) to a few hundred thousand dollars for more recent, larger IPs like Pokémon, significantly compressing traditional production costs.

Conclusion and Main Takeaways

The overarching advice for creators is to "get your hands dirty" and start creating. The key takeaways are:

• Start Creating: Begin generating content and putting "shots on target."
• Consume Viral Content: Analyze successful viral content to understand what resonates.
• Brand Palatability: Learn how to adapt viral concepts to be palatable and beneficial for brands.
• Build a Portfolio: Create a portfolio of "branded viral content" to "unlock the keys to the kingdom" and build a successful agency.
• Embrace Change: While the current workflow is effective, Sora's advancements will integrate and accelerate the entire process, making it faster and cheaper. The focus will shift to the quality of ideas and creative direction.

PJ Ace also recommends Ror Heath's GenHQ course ($99) for those seeking a comprehensive guide.