The argument in one line.
You can reverse-engineer any viral AI video by uploading it to ChatGPT or using a meta prompt generator to extract the exact prompt that created it.
Read if. Skip if.
- A content creator using AI video tools like Runway or C Dance who wants to quickly identify the prompts behind viral videos you admire.
- Someone experimenting with prompt engineering for AI video generation and looking for a systematic way to reverse-engineer successful outputs.
- A video creator with 0-6 months of AI tool experience who feels stuck because competitors don't share their prompts and wants a faster learning path.
- You're building AI video prompts from scratch and want to understand prompt design theory rather than copy existing successful ones.
- You're already proficient at prompt engineering and regularly create viral AI videos—this is entry-level reverse-engineering for you.
The full version, fast.
Any viral AI video can be reverse-engineered into its source prompt using ChatGPT, ending the gatekeeping that surrounds AI video creators. The basic method takes about thirty seconds: screen-record or download the video, drop the file directly into ChatGPT, and ask it to return the prompt that would recreate it, broken out by scene, camera, lighting, and audio. The upgraded method uses a meta system prompt, a prompt-generator that ingests the video and outputs both a faithful reproduction prompt and a V2 cinematic version with Hollywood-grade camera directions and motion design. Paste the result into Seedance, Higgsfield, or any video model, tag your own character, and you reliably reproduce the original's structure and style.
Chat with this breakdown.
Modern Creator members can chat with any breakdown — ask for the hook, quote a framework, find the exact transcript moment. Unlocks at T2: refer 3 friends + add your own API key.
Create a free account →Where the time goes.

01 · Hook + AI video showcase
Opens with viral AI clips over reaction-cam. Frames the problem: creators gatekeep their prompts. Promise: I end that today.

02 · Method 1 - The embarrassingly simple way
Screen share of ChatGPT. Drags video file in, types: This video was made in Seedance 2.0, give me a prompt that can recreate it. ChatGPT returns a multi-scene prompt with style, camera, lighting, and audio sections.

03 · Method 1 results + Magnific/FreePik recreation
Ports the extracted prompt into Magnific (FreePik), tags his AI character, generates a 10-second recreation. Side-by-side comparison shown. Declares Method 1 successful but basic.

04 · Bridge - teasing the better way
Calls Method 1 common sense. Teases the upcoming prompt generator as something that does it better AND delivers a V2 10x upgrade.

05 · Method 2 - AI Video Engine Database
Opens his Everything Bundle database. Shows the meta system prompt for Seedance 2.0 prompt generation. Notes all prompts are free on metricsmule.com/blog.

06 · Method 2 in ChatGPT - paste system prompt + upload video
Pastes full meta system prompt into ChatGPT. ChatGPT asks for video upload. Returns full video analysis with meta tags, camera tokens, motion descriptors, style keywords, then the full recreation prompt.

07 · Results + V2 cinematic upgrade
Shows recreated video side-by-side with original. Triggers V2 by typing yes to ChatGPT follow-up. Gets Hollywood camera directions, stronger realism, advanced cinematography.

08 · V2 generation in HiggsField AI
Pastes V2 prompt into HiggsField AI, changes hair color to blonde, tags his AI character. Generates and shows the cinematic upgrade result.

09 · CTA + outro
Subscribe reminder. Free prompt generator link in description. Fade to closing AI-generated clip.
Lines worth screenshotting.
- You can reverse-engineer any AI video by dropping the file directly into ChatGPT and asking for a prompt that recreates it.
- Most creators gatekeep their AI video prompts, which means systematic reverse-engineering is the fastest path to matching their output quality.
- A meta system prompt — a prompt that generates prompts — is more reusable than any single video prompt because it adapts to any input.
- Dragging a video into ChatGPT produces not just a recreatable prompt but a full breakdown of scenes, camera shots, lighting, and audio style.
- The V2 cinematic upgrade step — asking for Hollywood-level camera directions after the initial prompt — consistently produces better output than the first pass.
- Screen-recording a video you want to reverse-engineer is the workaround when right-click download is blocked on the platform.
- Meta tags and camera tokens extracted from a video analysis are reusable in future, unrelated AI video prompts — not just for duplicating the original.
The meta prompt is the product.
A prompt that generates prompts is a reusable research tool - one system prompt can reverse-engineer any video, forever.
- Build a meta system prompt for your own niche and package it as a one-time tool in MCN+
- Use the V2 upgrade conversation pattern in any AI workflow: deliver a V1, then offer a cinematic version with a single follow-up question
- The gatekeeping-buster hook works in any tutorial: name the thing they want, say you are giving it to them, immediately deliver
- Free blog post + paid database is the full funnel - the blog ranks for prompts, the bundle converts the buyers
- Tag your AI character in every generation so your brand is inside every output
Terms worth knowing.
- Reverse engineer (a prompt)
- Working backward from a finished AI-generated video or image to figure out the text prompt that likely produced it, so you can recreate or adapt the result.
- Seedance 2
- An AI video generation model (referred to in the transcript as 'c dance two') known for producing multi-shot videos from a single prompt.
- Multi-shot generated video
- An AI-generated clip that contains multiple distinct camera shots or scenes stitched together from a single prompt, rather than one continuous take.
- Meta system prompt
- A prompt designed to generate other prompts. You give it an input (like a video) and it outputs a finished, structured prompt you can paste into another AI tool.
- Prompt generator
- A reusable instruction set that turns rough inputs into polished, ready-to-use prompts for a specific tool or use case, such as AI video models.
- Magnific / Freepik
- An AI creative platform that bundles access to multiple image and video models behind one interface, letting users pick which underlying model runs a given prompt.
- AI character
- A consistent, reusable AI-generated person or avatar that creators tag into new prompts so the same face and look appear across different videos.
- Tagging a character
- Referencing a saved AI character inside a prompt (often with an @-style tag) so the generator uses that specific likeness instead of inventing a new one.
- Gemini Gems
- Custom, reusable mini-assistants inside Google's Gemini app, each loaded with its own instructions and context for a specific recurring task.
- Custom GPT
- A configurable version of ChatGPT preloaded with custom instructions, files, and tools, used to specialize the assistant for a particular workflow.
- Claude Project
- A workspace inside Anthropic's Claude app that bundles files, instructions, and chats so a recurring task has persistent context across conversations.
- Everything Bundle
- A paid product that packages a creator's full library of prompts, prompt generators, and databases together for a single price.
- Meta tags and tokens
- Keyword-style descriptors (camera shots, lighting terms, style cues) extracted from analyzing a video, useful as building blocks when writing new AI prompts.
- V2 cinematic upgrade
- An optional second pass that rewrites a base prompt with Hollywood-style camera directions, stronger realism, and richer motion and cinematography cues.
- Higgsfield AI
- An AI video generation platform focused on cinematic camera motion and stylized shots, often used to render prompt-driven short clips.
Things they pointed at.
Lines you could clip.
“The actual creator never shares what prompts were actually used. Well, today, that ends.”
“Give me a prompt that can recreate it. I mean, that's pretty embarrassingly simple and common sense.”
“Would you like a V2 cinematic upgrade with more advanced Hollywood-level camera directions, stronger realism, motion design, and basically more advanced cinematography?”
Word for word.
The bait, then the rug-pull.
The gatekeeping ends today. MetricsMule opens by playing the viral AI videos everyone is seeing then pivots hard: those creators never share the prompts. Two methods follow, one embarrassingly simple, one a full meta system that returns a V2 cinematic upgrade on command.
Named ideas worth stealing.
Meta System Prompt
A system prompt that takes any video as input and returns: multi-scene breakdown, style tokens, camera language, and a V2 cinematic upgrade. Reusable across any video.
V2 Cinematic Upgrade Loop
After getting the base recreation prompt, type yes to a follow-up question in the same ChatGPT thread to receive a Hollywood-grade rewrite with advanced camera directions.
Gatekeeping-Buster Hook
Open by naming the thing everyone wants but nobody shares, then immediately promise to share it. Creates instant authority and goodwill.
How they asked for the click.
“Check the links in the description for this prompt generator. Hit that subscribe button.”
Double CTA - subscribe + free prompt link. Product upsell (Everything Bundle) was done mid-video at ~4:15, not at the end, which is smart pacing.







































































