Ready for Production?

You've explored voice synthesis. Now understand what changes when moving to a production deployment serving accessible council content.

This is guidance, not implementation. Use this information to inform your accessibility strategy and technical planning.

Scale Considerations

The demo synthesizes on-demand. Production systems need caching strategies and batch processing for efficiency.

Content Volume
Content volume: pre-generate for static pages, real-time for dynamic
Caching
Caching strategy: hash-based keys prevent regeneration of unchanged content
Languages
Multi-language: consider Welsh voices for bilingual councils
Batch Processing
Batch processing: generate audio during off-peak hours for cost efficiency

Customization Options

The demo uses default settings. Production deployments are customized to your council's brand voice and accessibility requirements.

Customization Description Effort Estimate
Voice Selection Voice selection: choose voices matching your council's brand 1-2 days
SSML Templates SSML templates: create reusable templates for common content types 3-5 days
Pronunciation Pronunciation dictionary: maintain list of local terms and phonemes Ongoing
Player Styling Audio player styling: match to council website accessibility standards 2-3 days

Cost Comparison

Costs scale with content volume and engine choice. Here's a comparison between this demo and a typical production deployment (1M characters/month).

Service Demo Costs Production Costs (1M chars/month)
Polly Neural $40-100/month (1M chars/month)
Polly Standard $10-25/month (1M chars/month)
Amazon S3 $5-10/month
AWS Lambda $5-10/month
Total No cost to your council $60-145/month (neural) or $30-55/month (standard)
Note: Neural voices cost ~4x more but sound significantly more natural. Consider neural for public-facing content.

Security Checklist

The demo prioritizes ease of exploration. Production deployments need additional security measures for serving audio content.

Presigned URLs with short expiry (1 hour)
S3 bucket policy restricting direct access
Lambda function with minimal IAM permissions
CloudFront for high-traffic audio delivery
Content sanitization before synthesis
Rate limiting to prevent abuse

What's Your Next Step?

Are you ready to proceed with this scenario?