end-to-end viral ai video production workflow

Lead with recognizable public-domain IP so viewers feel the familiar inside the foreign.

Open with iconic scenes (Pompeii, Titanic, Marie Antoinette) that everyone knows, then layer absurd financial advice on top to create comedic contrast.

Why: Familiar references lower cognitive load and increase watch-time; the twist triggers the share reflex.

"we like recognizable IP to start so that people have the familiar with the foreign"

Comedy first, brand last prevents the AI-replacement backlash.

Make the ad feel like a Super-Bowl spot: ridiculous, joke-heavy, with only a short brand tie-in at the end.

Why: If people are laughing they watch to the end and don’t reach for pitchforks about job loss.

"we always try and lead... it's going to be comedy first, and there's going to be mostly jokes and a little bit of brand at the end"

Generate images before video to cut cost and iterate faster.

Script → ChatGPT shot list → still images in Rev → Veo3 frames-to-video → edit. This is 10× cheaper than blind text-to-video slot-machine burns.

Why: Images let you preview every frame for client approval and creative tweaks without burning expensive video credits.

"it's a lot easier and cheaper and faster to generate images... than it is to do everything text to video"

Use frames-to-video (image-to-video) to lock composition and character consistency.

Upload the exact still frame as the first frame to Veo3, then append motion prompts; the model respects the starting composition.

Why: Eliminates the ‘morphing face’ problem and gives director-level control over framing.

"instead of text to video, we're gonna do frames to video... that acts as our first frame"

Treat ChatGPT as a junior writer that sparks 1-in-100 good lines.

Prompt for historical moments + bad advice; iterate until one suggestion is funny or sparks a better human punch-up.

Why: AI is not funny on its own, but it’s an infinite idea machine that occasionally hands you gold.

"ChatGPT is not funny. 99 times out of 100, it's not funny, but one time it's good or it'll spark an idea"

Describe camera motion in plain English; the model understands either jargon or ‘move camera left’.

You can write ‘dolly in’ or simply ‘camera zooms to ship’; Veo3 interprets both.

Why: Lowers the skill floor so non-filmmakers can direct shots conversationally.

"you can also be stupid with it, like move camera left to end on the ship... it doesn't matter"

Stack best-of-breed models instead of one-suite-fits-all.

ChatGPT for scripting, Rev for photoreal images, Veo3 for talking characters, Kling/Luma/Minimax for non-dialogue scenes, Epidemic for music.

Why: Each model has a specialty; chaining them yields higher fidelity than any single platform.

"there's a lot of programs right now... VO3 is probably the best model at the moment for making characters talk"

Start with free spec spots in a niche you love, then expand.

Target beer, perfume, clothing—whatever excites you—create 3 spec ads, then pitch look-alike brands; use success to broaden portfolio.

Why: Passion sustains iteration; a focused niche portfolio converts faster than generic reels.

"follow your interests... you'll outlast your competition, because it feels like work you're following"