viral ai video creation workflow using sora 2 claude and perplexity
by @gregeisenberg
ABOUT THIS SKILL
distribution is the new moat; creators who master emerging platforms early gain unfair advantages. this workflow shows how to surgically craft short-form ai videos that maximize attention and startup distribution.
TECHNIQUES
KEY PRINCIPLES (10)
use unresolved tension as a hook to stop the scroll.
frameworks like 'contradiction hook' (e.g., 'I'm broke, but I'll show you my 50K MRR SaaS') exploit the brain's need to resolve cognitive dissonance.
Why: high-arousal emotions (FOMO, validation-seeking, pattern recognition) trigger dopamine and increase watch-time.
"the contradiction hook. this leverages the brain's attraction to unresolved tension."
early adopters of new tools capture outsized attention.
users who mastered youtube 2006-2012, instagram 2012-2019, and twitter 2019-2023 built massive media empires; sora 2 is the next frontier.
Why: algorithmic feeds reward novelty and reward creators who understand new formats before saturation.
"people who understood how to use youtube in 2006 to 2012 did better than people who didn't."
state the audience's core motivation before ideation.
for ideabrowser users, the deep desire is 'owning a SaaS that spits out cash'; every concept must map to this outcome.
Why: alignment between content promise and audience desire increases click-through and share rates.
"what is the main motivation for why people are signing up to your product?"
assign distinct roles to each ai tool.
perplexity acts as research agent, claude as strategist and prompt engineer, sora as production studio.
Why: specialized prompting reduces hallucination and produces higher-fidelity outputs tuned to each model's strengths.
"we're gonna use Perplexity for the research agent. we're gonna use Claude to write the Sora prompts."
score every concept against engagement criteria before production.
rate hook strength (first 3 seconds), pattern interrupts, emotional trigger, algorithm fit, and trend friendliness on a 1-10 scale.
Why: quantitative scoring removes guesswork and surfaces the highest viral ceiling ideas.
"rate 1 to 10 and briefly comment on hook strength... how likely is this concept to lead to high completion rate or shares?"
embed cinematic and audio cues directly in the prompt.
explicitly request synchronized speech, on-screen text, and visual elements like 'timer at 00:00:47' to maximize sora's capabilities.
Why: sora 2 can render synchronized speech and some text; precise prompting increases production value without reshoots.
"use quotes for the exact line I once spoken... describe the key visual elements... mention on-screen text explicitly"
treat every upload as data for the next prompt.
after publishing, return to claude with performance metrics and ask for refined prompts or style variations.
Why: continuous feedback tightens the resonance between content and audience, compounding reach over time.
"keep going back and forth and iterating... ask Claude to give you advice on how to prompt it."
master 10-15 second storytelling before formats expand.
early vine (6s) and tiktok (15s) winners dominated before longer uploads arrived; same window exists for sora 2.
Why: constraint breeds creativity; short formats force clear hooks and rapid payoff, training creators for future flexibility.
"reminder, Vine was six seconds back in the day. early days in TikTok was just 15 seconds."
WHAT'S INSIDE
This is a structured knowledge base — not a prompt file. Your AI retrieves principles semantically, understands the reasoning behind each technique, and connects to related skills via a knowledge graph.
Compatible with OpenClaw · Claude · ChatGPT
principles · semantic retrieval · knowledge graph
Free during beta · Sign in to save to dashboard