GreenLight

nspiration Creators typically lose money at two critical moments: right before they ship a video, and long after it goes live. Before shipping, forgotten sponsor clips, bad audio levels, or missing constraints can delay a launch or breach a contract. Long after a video is live, sponsor links and discount codes often expire or change—a phenomenon we call "Content Drift"—costing creators thousands in lost affiliate revenue. We wanted to build a "Creator Operations Protocol" that catches these costly mistakes at both ends of the publishing lifecycle.

What it does GreenLight is a two-part engine that secures a creator's content workflow:

Module 1 (Pre-Publish Readiness): Evaluates media files against readiness criteria before publishing. It runs FFmpeg to explicitly validate video constraints (resolution, aspect ratio) and uses Whisper AI transcription to ensure audio constraints are met (e.g., verifying a sponsor's name is spoken). Module 2 (Post-Publish Health & Drift): Monitors live content state. It calculates deterministic diffs between the local source-of-truth and the live YouTube data to detect "Content Drift" (like outdated sponsor links or codes). It then uses the YouTube Data API to safely patch and update the live video descriptions without unauthorized mutations. We also built a "Judge Mode" (/judge)—a 30-second deterministic cinematic presentation specifically engineered for hackathon judges using requestAnimationFrame and Framer Motion layout transitions.

How we built it We built GreenLight with a robust, feature-rich stack divided into a local processing backend and a premium web interface:

Backend: Python 3.9+ with FastAPI provides the core API. We used PostgreSQL and AsyncPG with SQLAlchemy and Alembic for fast, asynchronous database operations. The media processing engine leverages FFmpeg for video analysis and OpenAI Whisper for local audio transcription. Frontend: Built on Next.js 16 (Turbopack) and Tailwind CSS v4. We completely ditched the generic "Dark SaaS" look for a premium, editorial "Warm Beige" (#E8D8C3) aesthetic with Dark Brown (#241917) typography and Deep Red (#7A1D22) CTAs. 3D Artifacts: We used React Three Fiber to build an interactive 3D Hero Object (Hero3DObject.tsx) that represents the physical state of "Content Drift", featuring mouse parallax, warm key lights, and deep red emissive rims. Challenges we ran into One of the biggest challenges was ensuring zero unauthorized mutations when updating live YouTube videos. We had to build a cryptographically safe "Dry Run" architecture that calculates exact diffs of the video description so that we only update the specific sponsor links that drifted, without accidentally overwriting the rest of the creator's carefully formatted description. Getting the YouTube OAuth flow properly synced with our async database also required careful state management.

Accomplishments that we're proud of We are incredibly proud of the Judge Mode, which guarantees that anyone reviewing our project can see the exact value proposition in 30 seconds through a cinematic, choreographed UI sequence. We're also proud of successfully integrating hardware-intensive tools like FFmpeg and Whisper into a seamless, automated API pipeline, all wrapped in a visually stunning frontend.

What we learned We learned a lot about asynchronous database patterns in Python (FastAPI + AsyncPG), the intricacies of the YouTube Data API v3 (specifically the nuances of updating videos.snippet), and how to use React Three Fiber to elevate a standard dashboard into a premium, interactive experience.

What's next for GreenLight The next step is expanding our Pre-Publish engine to automatically detect and flag missing visual sponsor logos (using OCR), and expanding our Post-Publish engine to monitor a creator's entire back-catalog of thousands of videos, using cron jobs to automatically patch broken links across their entire channel history.