how i steal winning ads with ai - branded dropshipping

In this video, I take a winning ad from my own brand, PupRing, and completely recreate it using AI by retargeting it to completely different audience segments. By the end, we export a finished ad ready to launch made fully with ai. You can use this walkthrough on how to take a winning ad (yours or a competitor's) and break down the strategic thinking behind identifying which variables made the original ad work, which ones need to change, and how to rebuild the entire ad scene by scene. Tools Used: Analysis & Swipe File: Floh (waitlist: https://app.floh.ai/waitlist) Movement Extraction: Gemini (video → JSON) AI & Generation: Freepik (Nano Banana Pro model for image generation, Kling 3.0 for video generation) Editing: CapCut (final assembly, text overlays, music) Timestamps: 0:00 — Intro 0:15 — Extracting variables from a winning ad 0:41 — The strategy: same shots, different personas 2:52 — Analyzing the original ad & choosing Mother's Day 3:43 — Extracting key frames & winning variables 6:04 — Setting up workspace in Freepik 6:35 — Extracting motion data with Gemini 7:31 — Swapping & aging the character with AI 11:28 — Replacing character across all scenes 13:04 — Background swap to flower meadow 22:30 — Animating with Kling 3.0 + cost breakdown 32:15 — Refining prompts & regenerating clips 44:01 — Final passes & upscaling 49:00 — Stitching in CapCut + Mother's Day CTA 55:06 — Final export Key Points: • A winning ad isn't one-and-done — the same proven structure can be retargeted to entirely different audience personas using AI, multiplying one winner into many • Before touching any tool, separate your variables into two buckets: what made it work (keep these) vs. who it speaks to (change this). The controversial tattoo driving comments stays; the young girl becomes a mom • Most products have multiple buyer personas beyond the obvious — a dog mom jewelry brand where 90% of purchases are gifts means boyfriends, husbands, and kids are all untapped audiences for