- Home
- Skills
- Image & Video
- Reverse Any Video: Shot-by-Shot Recipe & Remake
More screenshots
Works with the AI Agents you already use
Reverse Any Video: Shot-by-Shot Recipe & Remake
Stop guessing why a video went viral or how an ad was edited.
Free
Reverse Any Video: Shot-by-Shot Recipe & Remake
Example session with this skill installed
Here is a 7-second product reel I like (reel.mp4). Reverse-engineer how it was edited and give me Remotion timings so I can make my own version for our new water bottle.
- Read your context and instructions
- Compiled the reverse any video
Video recipe: 7 s product reel
Formula: result-first hook, 3 macro details at 0.6 s, reveal, logo lock-up on the music drop.
Specs (measured with analyze_video.py)
| Aspect | Length | FPS | Shots | Cuts/min | Avg shot | First cut | Loudness |
|---|---|---|---|---|---|---|---|
| 9:16 | 7.0 s | 30 | 6 | 43 | 1.17 s | 1.2 s | -14 LUFS |
Shot list
| # | Time | Purpose | Camera | Device | Source |
|---|---|---|---|---|---|
| 1 | 0.0-1.2 s | Hook | ECU, push-in | Bold claim, 3 words | Measured + Observed |
| 2 | 1.2-2.0 s | Problem | MS, handheld | Caption per line | Measured + Observed |
| 3 | 2.0-2.6 s | Turn | Whip pan | SFX whoosh | Observed |
| 4 | 2.6-4.4 s | Proof | Macro x3 @ 0.6 s | Cuts on the beat | Measured |
| 5 | 4.4-6.0 s | Payoff | WS reveal | Result number | Observed |
| 6 | 6.0-7.0 s | CTA | Logo lock-up | Music drop | Observed |
Why it works
- The result is shown before the first cut at 1.2 s, so there is no slow intro.
- Shots under 1.2 s with a pattern change every 2 s keep retention high.
- The proof cuts land on the beat, so the fast edit feels intentional.
Remotion timings for your own version (new footage, your brand, royalty-free music)
const FPS = 30;
export const SHOTS = [
{ id: "hook", from: 0.0, dur: 1.2, text: "Ice cold. 24h." },
{ id: "problem", from: 1.2, dur: 0.8, text: "Warm water by noon?" },
{ id: "turn", from: 2.0, dur: 0.6, transition: "whip" },
{ id: "proof", from: 2.6, dur: 1.8, macros: 3 },
{ id: "payoff", from: 4.4, dur: 1.6, text: "-18 °C inside" },
{ id: "cta", from: 6.0, dur: 1.0, logo: true },
].map(s => ({ ...s, fromF: Math.round(s.from * FPS), durF: Math.round(s.dur * FPS) }));
Next steps: shoot 3 macro clips of the bottle (condensation, cap, ice), pick a 120 BPM royalty-free track so 0.6 s cuts land on half-beats, render 9:16 and 1:1.
Connects securely to your tools. The creator never sees your data.
About this skill
Stop guessing why a video went viral or how an ad was edited. This skill provides a technical framework to reverse-engineer any video into a reusable production recipe. It eliminates the guesswork in video analysis by using automated scripts to extract precise shot timings, pacing metrics, and color palettes, giving you the exact blueprint behind the creative.
What it does
- Automated cut detection using
analyze_video.pyto identify every shot transition and duration. - Metric extraction for cuts per minute, average shot length, first-cut timing, and loudness (LUFS).
- Visual profiling that generates contact sheets and shot frames to map framing, motion, and lighting.
- Structural decoding to identify hook patterns, retention devices, and beat maps across the timeline.
- Asset generation including AI video prompts, Remotion (React) code constants, or detailed production briefs.
How it works
- Analyze the local video file using the provided Python script to generate
analysis.jsonand a shot-by-shot contact sheet. - Profile each shot by documenting framing, camera motion, transitions, and typography styles against the measured data.
- Map the narrative structure into a beat map, grouping shots by their purpose like hooks, proof, or CTAs.
- Export a production-ready recipe using
templates/recipe.mdthat contains the technical specs and style rules.
Frameworks & tools
This skill requires ffmpeg and the pillow library for frame analysis. It optionally integrates with OpenAI Whisper for transcript extraction and supports output for Remotion (React) or AI video generators.
Why this beats prompting it yourself
Generic prompts often miss the mathematical pacing that makes video effective. By starting with hard data from analyze_video.py, you get millisecond-accurate cut lists and verified loudness levels that a standard LLM cannot see.
Use cases
- Breakdown competitor ads to understand their high-retention hook structures.
- Create AI video prompt sequences that maintain consistent style across multiple shots.
- Generate Remotion code constants to perfectly sync UI animations with a reference beat.
- Build internal style guides for editors to standardize the pacing of brand videos.
Known limitations
Requires local video files rather than URLs to respect platform protections. Requires Python, FFmpeg, and Pillow installed on the local environment for the measurement script to function.
Requires local Python environment with FFmpeg and Pillow. Cannot bypass platform download protections for URLs. Analysis precision depends on cut-detection threshold settings.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
1 use
Times used to date
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 6 days ago
- Free with an Agensi account