Skip to content
1st Place

AdSuperSense

AdSuperSense: The Next Generation of YouTube AdSense Problem: Standard mid-roll ads break video flow, causing viewer frustration and churn.

Code Video Cerebral Valley

Core trick

Extracts the frames around a video hotspot and generates a bridging clip with product compositing via image and video models.

Tags

visioncreative/music

Description

AdSuperSense: The Next Generation of YouTube AdSense Problem: Standard mid-roll ads break video flow, causing viewer frustration and churn. Solution: We eliminate jarring interruptions by blending ads seamlessly into natural transitions using Nano Banana 2 and Google Veo. How It Works: - Smart Hotspots: Our algorithm identifies natural sequence breaks. Creators select a hotspot and assign an advertiser. - Frame Extraction: We isolate the final frame of the current scene and the first of the next. - Product Integration: Nano Banana 2 (Gemini 3 Flash Image) composites the product perfectly into the frame. - AI Video & Audio: Google Veo generates smooth video bridging the scenes with the product composition, plus native audio for the ad script that matches the scene's ambiance. Result: Contextual, on-demand ads that preserve immersion, maintain flow, and maximize advertiser value.
Min0laa/Ad-SuperSense★ 00 forksTypeScriptCreated Mar 14, 2026 · last push Mar 16, 2026

README excerpt

# Gemini-3-Paris-Hackathon

AI-powered YouTube ad injection pipeline. Finds peak-energy moments in any YouTube video and generates a frame-accurate handoff package for AI ad rendering.

---

## Architecture

```
browser-extension/          Chrome Extension (Plasmo + React)
  contents/                   Injects UI into YouTube Studio upload flow
  assets/                     Static brand assets

app/                        Node.js / TypeScript engine
  src/lib/placement.ts        Scoring engine — finds injection points
  src/lib/gemini-service.ts   Gemini API bridge (mock → real)
  handoff.ts                  CLI + exportable analyzeVideo() function
  server.ts                   Express server (port 8080)
  main.py                     FastAPI server (port 8000) ← main backend
  outputs/                    Generated frames + handoff.json (gitignored)

backend/                    Separate video analysis service
  src/main.py                 FastAPI hotspot detector (file upload API)
```

---

## Prerequisites

- **Node.js 18+** and npm
- **Python 3.8+**
- **ffmpeg** + **yt-dlp** — `brew install ffmpeg yt-dlp`
- **FastAPI + uvicorn** — `pip install fastapi uvicorn`

---

## 1 — FastAPI Handoff Server

Serves injection spots and frame images to the Chrome Extension.

```bash
cd app
uvicorn main:app --reload --port 8000
```

| Route | Description |
|---|---|
| `GET /api/spots/{video_id}` | Analyze video or return cached result |
| `GET /outputs/{video_id}/spot_N/frame_a.jpg` | Extr

Team