This AI Creative Agent clones 36 Million-view Ads on Autopilot (Kling2.6 x Gemini 3 Airtable System)

TLDR
This video introduces an AI agent system built on N8N that automates the repurposing of successful, viral ads. By utilizing Gemini 3 for scene analysis and Nano Banana for generation, the system rebuilds proven content formats featuring custom products and characters without manual prompting. The workflow allows creators to identify winning ads, automate the entire production process, and maintain quality control, significantly speeding up content creation.

Summary
The video introduces an AI creative agent designed to automate the repurposing of successful viral ads and content. It addresses the inefficiency of starting content creation from scratch, asserting that attention follows patterns and winning formats can be scaled. The core system utilizes Gemini 3 for video analysis, breaking content into scenes, then rebuilding new versions featuring specific products, characters, and creative directions, all without manual prompting. The presenter aims to teach viewers how to identify high-performing ads for their brand or clients and leverage this AI agent system to launch proven content formats in an automated way.

To find winning ads or content, the video suggests several methods. For competitive ad intelligence, it recommends using Magic Brief (a third-party tool) to filter ads by industry, theme, brand, or language. Users can identify "winners" based on Magic Brief's proprietary performance estimates (summarizing over 60 attributes like script quality and ad longevity) or by filtering for ads that have been running for an extended period (e.g., over 90 days), indicating continuous investment and likely success. For finding viral content on social media like Instagram, the video suggests manually browsing reels for high views or using a Chrome extension like "Sort Feed for Instagram" to quickly sort posts by viewership.

The AI creative agent system is built on N8N, a no-code automation platform. The process begins by uploading a reference video, character images, product images, and a creative request into an AirTable database. Gemini then analyzes the reference video, detailing its scenes using a "SEAL CAM" framework (Subject, Environment, Action, Lighting, Camera, Meta tokens) and generating detailed visual prompts. These prompts, along with specific creative directions, are then fed into Nano Banana Pro (accessed via Wavespeed, an AI model aggregator) to generate starting images without watermarks. Finally, these images and video prompts are used with Cling 2.6 (also via Wavespeed) to produce the final video clips for each scene. As a bonus, Suno AI (via Key.AI's API) can generate background music.

This comprehensive system automates the entire content generation workflow, from analysis and prompting to image, video, and music creation, often in a single click. The video demonstrates this with several examples, including repurposing a Prada ad with custom apparel and a jewelry ad with custom characters and products, showcasing the agent's ability to adapt styles and elements while maintaining the original ad's winning format. While automation significantly speeds up creation, the system also allows for manual inspection of prompts, images, and videos at each stage, ensuring quality control and incorporating human creativity. The presenter concludes by highlighting the benefits of fast-tracking content creation with proven formats and invites viewers to explore the RoboNuggets community for further AI training and resources.