Back to projects

PX47

Multimedia binary processing engine

Scalable backend pipeline utilizing ffmpeg and audiowaveform for high-throughput stream analysis. S3-integrated with pre-signed upload flows.

PX47 Demo

View on GitHub

Tech Stack

Node.jsExpressffmpegAudioWaveFormAWS S3Docker

Overview

This project is a backend data-processing pipeline for audio and video files, built with Express.js. Users can upload files directly to S3 using pre-signed URLs, and files are processed asynchronously using BullMQ + Redis workers. The pipeline handles format conversion, metadata extraction, waveform generation, and uploads the processed files back to S3.

Architecture Overview

Video and Audio Processing Architecture

Features

  • Direct S3 uploads via pre-signed URLs
  • Automatic worker triggering after file upload
  • Audio format conversion (e.g., MP3/AAC)
  • Metadata extraction from audio/video files
  • Waveform generation stored as JSON
  • Processed files and waveform JSON are uploaded back to S3
  • Asynchronous processing with BullMQ + Redis

Workflow

  1. Client requests a pre-signed URL from the backend
  2. Client uploads the file directly to S3
  3. A worker picks up the uploaded file
  4. Worker downloads the file locally (required by FFmpeg)
  5. Extracts metadata using FFprobe
  6. Generates a mute video from the original video
  7. Extracts audio-only track and converts to AAC format
  8. Generates waveform JSON for the audio
  9. Uploads processed files back to S3
  10. Saves all metadata and S3 keys in the database