Media - GPU-Accelerated Video Transcoding Service

A video transcoding service built with Rails that leverages NVIDIA CUDA hardware acceleration to transcode videos to multiple quality levels. The service processes videos asynchronously using background jobs and stores outputs to cloud storage.

Getting Started

Prerequisites

• Ruby 3.4.7
• FFmpeg with CUDA support
• NVIDIA GPU with NVENC/NVDEC support
• Docker & Docker Compose

Setup

1. Clone the repository

   git clone <repository-url>
   cd media

2. Install dependencies

   bundle install

3. Start services and setup database

   docker-compose up -d
   bin/rails db:create
   bin/rails db:prepare
   bin/rails data:migrate

4. Start the application

   bin/dev

5. Access the application

   • API: http://localhost:3009
   • Sidekiq UI: http://localhost:3009/sidekiq

Key Features

• GPU-Accelerated Transcoding - NVIDIA CUDA hardware encoding/decoding
• Multiple Quality Profiles - Automatic transcoding to 480p, 720p, and 1080p
• AV1 Encoding - Modern AV1 codec with hardware acceleration
• Asynchronous Processing - Background job processing with Sidekiq
• Cloud Storage - S3 and IBM Cloud Object Storage integration

GPU Compatibility

This service requires an NVIDIA GPU with AV1 NVENC support (RTX 40 series or newer). See the NVIDIA Video Encode and Decode GPU Support Matrix for compatible hardware.

FFmpeg is compiled from source with CUDA support. To build the FFmpeg image locally:

docker build -f utils/ffmpeg/Dockerfile -t ffmpeg-cuda:latest .

Testing

bundle exec rspec

License

Apache License 2.0