Cloud Transcriber
Web application for media transcription using AWS, Whisper ASR, and Node.js. Scalable and secure architecture.
Jul 2024 - Nov 2024 • 4 months
Tech Stack
AWSNginxExpressNode.jsTerraformCloud ComputingMicroservicesWhisper ASRFFmpegPostman
This project details a web application designed to allow users to upload media files (audio and video) and generate transcriptions of their audio content. The application leverages various AWS services and open-source technologies to ensure efficient processing, scalability, and security.
Application Overview
The core functionality of the application involves:
- Converting video files to MP3 audio using FFmpeg during the upload process.
- Transcribing audio files to text using the open-source Whisper ASR (Automatic Speech Recognition) service.
- Automatically scaling the transcription service based on workload.
- Storing and allowing users to retrieve and view transcriptions from a database.
Application Architecture
The application is deployed on AWS in the ap-southeast-2 region. Key components and their roles include:
- Route 53: Maps the custom domain to the API Gateway endpoint.
- API Gateway: Routes HTTPS traffic to the EC2 instance.
- EC2 Instance: Hosts Docker containers running an Express backend (for REST API) and Nginx (for web client and reverse proxy).
- ElastiCache: Provides Memcached-based caching to improve API response times.
- ECS Cluster (ECS Whisper Service): Hosts and auto-scales the Whisper transcription service based on the total duration of pending transcription tasks in the SQS queue.
- Lambda Functions:
- Conversion Service: Extracts audio from media uploaded to S3 using an FFmpeg layer and adds duration to metadata.
- Custom Metric Calculator: Periodically calculates the total queued audio duration for transcription and publishes it to CloudWatch.
- S3 Buckets:
- Media Bucket: Stores media files uploaded by users.
- Audio Bucket: Stores MP3 audio files extracted from uploaded media.
- Transcription Bucket: Stores generated transcription text files.
- SQS Queues:
- Audio Processing Queue: Notifies the server when audio extraction completes.
- Transcription Jobs Queue: Manages the queue of transcription jobs.
- Cognito: Handles user authentication with Google OAuth integration.
- CloudWatch: Monitors logs, schedules Lambda functions, and triggers ECS scaling policies.
- Secrets Manager: Securely stores DocumentDB credentials.
- Parameter Store: Stores configuration data.
- DocumentDB: Serves as the database for user audio files and transcription data.