🎥 VDO-Process — Media Processing & Normalization MVP

A hackathon MVP for uploading, processing, normalizing, merging, and streaming large video files using an async, scalable backend architecture.

This project demonstrates how heterogeneous video assets (different resolutions, FPS, audio levels, sources) can be analyzed, normalized, merged, and streamed through a unified API.

---

🚀 What This Project Does

• Upload large video files safely (no backend memory usage)
• Analyze video/audio metadata (FFprobe)
• Normalize video & audio to a target profile
• Merge multiple videos (with transitions)
• Track processing jobs asynchronously
• Stream processed videos directly to the frontend

All processing is done asynchronously using a job queue and background workers.

---

🧠 Key Features

• Large-file safe uploads using signed URLs
• Async job processing with Redis queue
• Video normalization (resolution, FPS, audio loudness)
• Video merging (crossfade support)
• Streamable outputs via object storage
• Hackathon-ready MVP architecture

---

🏗️ Architecture Overview

Frontend (React)
   ↓
FastAPI Backend
   ├── Signed upload URLs
   ├── Job orchestration
   ├── Job status APIs
   ↓
Redis (Job Queue)
   ↓
Worker Process
   ├── FFprobe (analysis)
   ├── FFmpeg (normalize / merge)
   ↓
MinIO (Object Storage)
   ├── raw videos
   ├── processed videos
   └── previews

---

🛠 Tech Stack

Backend

• FastAPI
• Python
• FFmpeg / FFprobe
• Redis (job broker)
• MinIO (S3-compatible object storage)

Frontend

• React

Infrastructure

• Docker & Docker Compose

---

▶️ How to Run Locally

1️⃣ Start infrastructure

docker compose up -d

Create a bucket in MinIO named:

media

---

2️⃣ Start backend

cd backend
uvicorn main:app --reload

API docs:

http://localhost:8000/docs

---

3️⃣ Start worker

python worker.py

---

📡 Example API Endpoints

• POST /assets/upload-url → get signed upload URL
• POST /create-job → create processing job
• GET /get-job-status/{job_id} → job progress
• GET /stream/{asset_id} → stream video

---

⚠️ MVP Notes

This is a hackathon MVP, not a production system.

• In-memory job state
• Single worker
• No auth
• No retries or persistence guarantees

---

🎯 Why This Matters

This project demonstrates:

• Practical handling of large media files
• Async processing pipelines
• Real-world media tooling (FFmpeg)

---