Show HN: Serverless video processing with AWS Lambda

4 months ago 4

🎥 Watch the Demo Video - See Auto-Vid in action with live AWS Lambda demonstration

A production-ready serverless video enrichment pipeline that uses a declarative JSON format to automatically add AI-powered TTS, music, and sound effects to video.

🎬 Professional Video Processing - MoviePy-powered video editing with precise timeline control
🗣️ AI Text-to-Speech - AWS Polly with 90+ voices, multiple engines, and SSML support
🎵 Smart Audio Mixing - Background music with crossfading, ducking, and volume control
🔔 Webhook Notifications - Real-time job completion notifications with retry logic
☁️ Managed S3 Storage - Automatic bucket creation with organized asset management
🔐 API Security - API key authentication with rate limiting (2 req/sec, 50/day)
📊 Scalable Architecture - SQS queuing, Lambda concurrency, and retry logic

🔧 AWS Lambda Implementation

AWS Lambda Usage in Auto-Vid:

Three Lambda Functions Architecture:

Submit Job Lambda (Python Runtime)
- Validates incoming JSON job specifications using Pydantic models
- Stores job metadata in DynamoDB with "submitted" status
- Queues processing jobs to SQS for reliable delivery
- Returns comprehensive job information via API Gateway
Video Processor Lambda (Container Runtime)
- Processes SQS messages containing video job specifications
- Downloads video/audio assets from S3 to Lambda's /tmp storage
- Generates AI speech using AWS Polly integration
- Performs complex video editing with MoviePy (audio mixing, ducking, crossfading)
- Uploads final processed videos back to S3
- Generates pre-signed S3 URLs for secure video downloads
- Updates job status in DynamoDB and sends webhook notifications
Get Status Lambda (Python Runtime)
- Retrieves job status and metadata from DynamoDB
- Returns comprehensive job information via API Gateway

Lambda Container Benefits:

Handles large video processing libraries (MoviePy, FFmpeg)
Optimized Docker image (360MB) for faster cold starts
Scales automatically from 0 to hundreds of concurrent video processing jobs
Pay-per-use model - zero cost when idle, cost-effective at scale

AWS CLI configured with appropriate permissions
SAM CLI installed
Python 3.12+

# Verify your setup aws sts get-caller-identity aws polly describe-voices --region us-east-1

# Clone and build git clone https://github.com/ossamaweb/auto-vid.git cd auto-vid sam build # Takes time to build the video processor Docker image # Phase 1: Deploy without authentication sam deploy --guided --parameter-overrides DeployUsagePlan=false # Answer 'Y' to create managed ECR repositories # Note the API URLs from output # Phase 2: Add API key authentication sam deploy --parameter-overrides DeployUsagePlan=true # After deployment, sync demo assets to S3 # Replace with your actual S3 bucket from deployment output BUCKET_NAME="your-bucket-name" aws s3 sync ./media/assets/ s3://$BUCKET_NAME/assets/ aws s3 sync ./media/inputs/ s3://$BUCKET_NAME/inputs/ # Update sample job files with your bucket name # Option 1: Automatic update (cross-platform) perl -i -pe "s/your-bucket-name/$BUCKET_NAME/g" samples/production/*.json # Option 2: Manual update # Edit samples/production/*.json files and replace "your-bucket-name" with your actual bucket name

# Replace with your actual API URL from deployment output API_URL="https://your-api-id.execute-api.us-east-2.amazonaws.com/Prod" # Get API key from AWS Console: API Gateway → API Keys → auto-vid-api-key-{stack-name} → Show API_KEY="your-actual-api-key-from-aws-console" # Submit test job (API key required) curl -X POST $API_URL/submit \ -H "Content-Type: application/json" \ -H "X-API-Key: $API_KEY" \ -d @samples/production/00_api_demo_video.spec.json # Check status (replace JOB_ID with actual job ID) curl $API_URL/status/JOB_ID \ -H "X-API-Key: $API_KEY"

{ "jobInfo": { "projectId": "api_demo", "title": "API Test" }, "assets": { "video": { "id": "main_video", "source": "s3://your-bucket-name/inputs/api_demo_video.mp4" }, "audio": [ { "id": "track", "source": "s3://your-bucket-name/assets/music/Alternate - Vibe Tracks.mp3" } ] }, "backgroundMusic": { "playlist": ["track"] }, "timeline": [ { "start": 4, "type": "tts", "data": { "text": "Welcome to Auto-Vid! A serverless video enrichment pipeline.", "duckingLevel": 0.1 } } ], "output": { "filename": "api_demo_video.mp4" } }

# Setup cp .env.example .env # Add your AWS credentials python3 -m pip install -r requirements.txt # Test components locally python3 test_tts_local.py english python3 test_local.py # Deploy for full testing (requires AWS) sam build # Takes time to build the video processor Docker image sam deploy --guided

├── src/ │ ├── submit_job/ # Job submission API │ │ ├── app.py # Lambda handler │ │ └── requirements.txt │ ├── get_status/ # Status checking API │ │ ├── app.py # Lambda handler │ │ └── requirements.txt │ └── video_processor/ # Core video processing │ ├── app.py # Lambda handler │ ├── video_processor.py # Main video processing logic │ ├── asset_manager.py # S3 integration │ ├── tts_generator.py # AWS Polly integration │ └── webhook_notifier.py # Webhook notifications ├── layers/shared/ # Shared code between Lambda functions │ ├── job_spec_models.py # Pydantic models for job specification │ ├── job_validator.py # Validation functions │ ├── job_manager.py # DynamoDB operations │ ├── response_formatter.py # Standardized responses │ ├── polly_constants.py # Voice/language definitions │ └── requirements.txt # Shared dependencies ├── media/ # Media files (matches S3 structure) │ ├── assets/ # Demo audio files (music + sfx) │ ├── inputs/ # Sample input videos │ └── outputs/ # Generated videos (gitignored) ├── tmp/ # Temporary files during processing (gitignored) ├── docs/ # Documentation ├── samples/ # Example job specifications ├── template.yaml # SAM infrastructure ├── Dockerfile.videoprocessor # Container definition ├── test_*.py # Local testing scripts ├── .env.example # Environment template └── env.json.example # Container env template

Notes:

Local testing only works for individual components (TTS, video processing, S3 upload and webhooks)
Full integration testing requires AWS deployment (SQS + DynamoDB)
SAM handles container build, ECR management, and infrastructure automatically
Use sam deploy for updates after initial guided setup

Video Processing Performance:

Uses 3008MB memory (compatible with all AWS accounts)
For higher performance, request Lambda memory quota increase via AWS Support

# Delete the entire stack and all resources (replace with your actual stack name) aws cloudformation delete-stack --stack-name <your-stack-name> # Wait for deletion to complete aws cloudformation wait stack-delete-complete --stack-name <your-stack-name> # Verify deletion aws cloudformation describe-stacks --stack-name <your-stack-name> # Should return: "Stack with id <stack-name> does not exist"