how i steal viral videos with ai - branded dropshipping
This is the exact workflow I use to identify a viral video, extract its core variables — movement, framing, aesthetic — and recreate it for my own product using AI. No original footage. No film crew. Just a screenshot, prompt, and a few image and video models. By the end you'll know how to clone any viral video for your brand and hand the entire process off to a VA.
Tools Used:
Analysis & Swipe File: Floh, OpenClaw Agent (floh waitlist: https://app.floh.ai/waitlist)
Movement Extraction: Gemini (video → JSON)
Image Generation: Freepik, Nano Banana Pro
Video Generation: Google Veo 3, Kling 3.0, Sora 2 Pro
Timestamps:
00:00 – Intro
00:10 – Finding a viral video and saving it to your swipe file
01:50 – Screenshotting the first frame of each scene
02:33 – Extracting scene movement as JSON using Gemini
04:04 – Setting up your generation workspace in Freepik
05:06 – Removing text and generating the background image (Nano Banana Pro)
07:21 – Running 10 image variations and picking the best
09:50 – Upscaling the winner 2x before video generation
10:12 – Generating video from your image: Veo 3 vs Cling 3.0 vs Sora 2 Pro
14:10 – Comparing outputs and picking the winner
16:02 – Adding text + music overlay — done
Key Points
- The real unlock is extracting scene movement as JSON — this becomes the brief for video generation so the AI recreates the camera behavior exactly, not just the look.
- Generate the background image first without text, then add copy in post — noticeably better output quality.
- Run 10 image variations before moving to video. By the time you're at video gen you should already know what you want — 2 variations is enough.
- Kling 3.0 produced better motion than Veo 3 at roughly half the credit cost. Don't default to the expensive model without testing.
- Sora 2 Pro hallucinated completely on this test — worth knowing before burning credits.
The entire workflow is fully VA-executable (and ai agent executable hehe ). Design it once, hand it off, get content
Watch on YouTube ↗
(saves to browser)
Sign in to unlock AI tutor explanation · ⚡30
More on: Prompt Craft
View skill →Related AI Lessons
⚡
⚡
⚡
⚡
What makes an AI image workflow useful for real commercial output?
Dev.to AI
How to Write Better AI Image Prompts for Midjourney (With Examples That Actually Work)
Medium · ChatGPT
Image to Video AI: The Complete Workflow Playbook That Actually Produces Results
Medium · AI
Image Harvest v1.0.2: Internationalization, Free Pro Trial & Quality-of-Life Improvements
Dev.to · kyriewen
Chapters (11)
Intro
0:10
Finding a viral video and saving it to your swipe file
1:50
Screenshotting the first frame of each scene
2:33
Extracting scene movement as JSON using Gemini
4:04
Setting up your generation workspace in Freepik
5:06
Removing text and generating the background image (Nano Banana Pro)
7:21
Running 10 image variations and picking the best
9:50
Upscaling the winner 2x before video generation
10:12
Generating video from your image: Veo 3 vs Cling 3.0 vs Sora 2 Pro
14:10
Comparing outputs and picking the winner
16:02
Adding text + music overlay — done
🎓
Tutor Explanation
DeepCamp AI