Clone Any Viral Reel, Frame for Frame (the AI Does the Edit)
I take a reel that already went viral, decode exactly why it works, and rebuild it on my own footage frame for frame. The AI does the reverse-engineering and almost all of the editing. I packaged the whole thing into a single Claude Code skill so it is basically one command, but below is the manual version with every prompt, so you can run it yourself.
What you're getting
A repeatable system for turning any viral reel into your own version. Decode any reel from its link: the spoken script, the on-screen text, the cuts, the music, and the real reason it went viral. A frame by frame shot list: every shot you need to film, in order, with the framing, the action, the length, and the exact text that lands on each one. An AI edit pass that assembles your clips into the same video: same color, same cuts on the beat, words that sit behind your head, the same transitions and the same music. You film. The AI decodes, plans, and edits.
The one idea
You are not copying the video. You are copying the DNA. A reel goes viral because of structure: the first two seconds, the rhythm of the cuts, where the text lands, when the music drops. That skeleton is what you take. The footage, the words, and the topic stay yours. Decode the skeleton, rebuild it with your own body.
Step 1: Decode the reel
Drop the link into Claude Code and let it pull the video apart. It scrapes the reel, transcribes every spoken word with exact timestamps, looks at the video every couple of seconds to read what is on screen, and pulls the caption, the hashtags and the view count. Then it writes you a report: the hook word for word, the structure beat by beat, the on-screen text, the music, the call to action, and which lever actually drove the views.
Decode this reel and tell me why it went viral, then break down everything I need to rebuild it: PASTE_REEL_LINK_HERE
Scrape the video, transcribe the spoken script with timestamps, read what is on screen every 2 seconds, and lay out the hook, the structure beat by beat, the on-screen text, the music, and the call to action.Step 2: Get the shot list
Next it turns that decode into a build plan. Not a vibe, an actual list. Every single shot you need to film, in order, with the camera framing, what you do on camera, how long it runs, and the exact words that land on it. It even separates the shots you film from the moments built in the edit, like the flashes, the split-screen and the title card. You shoot straight off the list.
Now turn that decode into a frame by frame shot list I can film from. For every shot give me: the framing, what I do on camera, how long to roll, and the exact text that goes on it. Mark which shots I film and which are built in the edit.Step 3: Film your clips
This is the only human part. You read the shot list and film each clip on your phone. Same wardrobe, same lighting, same poses the plan calls for. Drop every clip into one folder and hand it back. If you shot it to match the plan, the edit drops in clean.
Step 4: Let the AI edit it
This is where it gets stupid. Point the AI at your folder and it builds the whole edit. It transcribes your clips and reads the framing of each one. It cuts you out of the background with pixel-level masks, so the big words sit behind your head instead of covering your face, and your eyes can glow. It matches the color grade, lines the cuts up on the beat, conforms your slow motion, and drops in the flash, the split-screen and the strobes. It rebuilds the soundtrack: it splits the music off the original audio and lays the transitions and the music hits exactly on your cuts. Then it reviews its own video frame by frame, checks the lip sync and the loudness, and fixes whatever is off. On repeat, until it is right.
Here are my clips in this folder. Transcribe them, study the framing of each, and edit them into a frame for frame recreation of the reel we decoded.
Match the grade, cut on the beat, put the big words behind my head, copy the transitions and the music, then review your own output frame by frame and fix anything that is off.The full tool stack
Everything the AI reached for, and the job it did. Claude Code: the brain that ran all of it and made the calls. Apify: pulled the reel and its data off Instagram. Whisper: transcribed every spoken word with exact timing, for the original and for your clips. GPT-4o Vision: looked at the frames and described the framing, the poses, and the on-screen text. SAM2: traced you out of every frame so text could sit behind your head and your eyes could glow. FFmpeg: every cut, color grade, slow motion, flash, split-screen, caption and the final mix. Demucs: split the music away from the voice so the soundtrack could be rebuilt on your cuts. Free tools, one operator, no editor.
The part nobody else does
The reason it comes out clean is that the AI grades its own homework. After every build it pulls frames out of its own video and actually looks at them, measures the loudness across the whole track, checks that your lips match the audio, and confirms the music hits land on the cuts. When something is off it changes one number and rebuilds. That self-review loop is the difference between an AI toy and a real edit.
One rule
Copy structure, not people. Use your own face, your own footage, your own topic. Decoding why a video worked is fair game. Lifting someone's clips or faking a real person is not. Build your version.
Get the next one first
New prompts every week.
Free. The new drops and the tools behind them, before they hit the feed.
No spam · New issues Sunday · Unsubscribe anytime
Need it custom?
Want this built for you?
Tell me the idea and I’ll build it. An app, a tool, an automation. You don’t need to be technical.