233M Views in 3 Days: The David Beckham AI Workflow

What It Covers

PJ Ace reveals his complete AI video production workflow that generated 233 million views for a David Beckham ad in three days. He demonstrates the five-role team structure, specific tools like Ideogram and Kling 2.6, and the two-by-two grid technique for maintaining visual consistency across scenes in AI-generated commercials.

Key Questions Answered

• •Two-by-Two Grid Workflow: Generate images in four-shot sequences using a two-by-two grid format to maintain consistent lighting, characters, and locations across all frames. Crop individual shots from the grid, upscale them, then animate each frame separately in sequence. This technique eliminates the challenge of matching lighting and visual continuity between disconnected shots.
• •Five-Role Production Structure: Split AI video production into writer, director, cinematographer, animator, and editor roles mirroring traditional animation pipelines. This team approach scales beyond solo creators and allows preexisting filmmakers to collaborate effectively. The IMA David Beckham ad used this structure with multiple cinematographers and animators working from director-provided reference images and shot specifications.
• •Image-to-Video Over Text-to-Video: Upload reference images to video generation tools rather than relying on text prompts alone. Use the last frame of one clip as the first frame of the next for smooth transitions. Upload reference images to ChatGPT to extract visual language prompts that can recreate similar aesthetics without legal issues from using copyrighted movie stills.
• •Cost and Timeline Transformation: AI video production costs one-tenth of traditional commercials, bringing million-dollar quality down to six-figure budgets or lower. Entry-level AI filmmaking agencies charge between ten thousand and thirty thousand dollars per ad. Production timelines compress from four to twelve weeks down to days, with the Kalshi NBA Finals ad completed in two days.
• •Kling 2.6 Motion Control Technique: Record a driving performance video of an actor, generate a single reference image of the desired character or setting, then upload both to Kling 2.6 motion control. The tool retrofits the actor's facial performance and movements onto the generated character, enabling one actor to play multiple roles with different voices using Eleven Labs voice cloning.