ADIFY

Inspiration

The inspiration for Adify comes from the "creative wall" faced by millions of independent creators and local artisans. We saw talented entrepreneurs pouring their hearts into products but feeling invisible in a digital world dominated by big-budget agencies. For them, professional marketing isn’t just expensive—it’s an intimidating barrier that makes them feel they don’t belong. We watched creators struggle with complex design software or settle for blurry screenshots because they couldn't afford a designer. Adify was born to break this wall, turning a simple product link into a professional narrative so every creator, regardless of budget or skill, has the power to be seen and heard.

What it does

Adify (ProductAd.ai) is an AI-powered creative assistant that automates the entire ad generation workflow:

URL-based Scraping: Instantly extracts product data from any e-commerce or product website.

Visual Heat-Mapping & Scoring: Analyzes the generated ad to predict eye-tracking patterns and scores the creative based on design principles like contrast, balance, and readability.

Video-to-Thumbnail Engine: Automatically "scrubs" product videos to identify and extract the highest-quality "hero" frames for the ad background.

AI Copywriting: Generates human-like headlines, taglines, and product-specific CTAs while avoiding generic phrases.

Adobe Express Integration: Intelligently places content onto visually optimized templates and renders the final ad directly inside the Adobe Express canvas.

How we built it

We developed a hybrid architecture combining generative AI with a deterministic layout engine to balance creativity with structural reliability:

Backend: Built with FastAPI for high-performance asynchronous processing and Firecrawl for AI-powered web scraping. We integrated FFmpeg to handle the heavy lifting of video processing.

Frontend: Developed as an Adobe Express Add-on using React and the HTML5 Canvas API for precise, pixel-level text placement.

AI Layer: Leveraged the Google Gemini API for sophisticated text generation and Gemini Vision for semantic image/video analysis.

Challenges we ran into

The "Blurry Frame" Problem: Static thumbnails from videos are often low-quality. We solved this by creating a scoring pipeline where the AI analyzes multiple candidate frames for high product visibility and lighting contrast.

Data Sanitization: Scraped titles often arrived with HTML entities (like &). We implemented robust backend sanitization to ensure copy remained professional.

Pixel-Perfect Logic: We moved all color sampling and typography scaling to frontend canvas logic, ensuring consistent readability and design-rule compliance.

Accomplishments that we're proud of

We successfully built a system that turns a messy product link into a professional design in seconds. We are particularly proud of our Deterministic Rendering Engine. Unlike pure AI generation which can be unpredictable, our engine ensures that no matter how long the copy is, it is always styled and scaled perfectly relative to the Visual Heat-Map requirements.

What we learned

We learned the true power of hybrid intelligence. While AI is unmatched for creative tasks, traditional deterministic code remains essential for reliable layouts. We also gained deep technical insights into the Adobe Express SDK regarding complex asynchronous flows.