Case Study: Bandwidth achieves reliable, low-latency, cost-saving text-to-speech with Amazon Polly

Building a Reliable Text-to-Speech Service with Amazon Polly

Bandwidth, a communications CPaaS provider that powers voice and messaging for apps and enterprises, needed a more reliable, natural-sounding, and cost-effective text-to-speech solution after uptime and caching limitations with their previous vendor. Their customers rely on TTS for menus, reminders, and order information, so Bandwidth sought a fast, voice-rich engine that could handle dynamic text interpretation and support caching.

Bandwidth integrated Amazon Polly as its primary TTS engine, building an internal service that checks an S3 cache, streams Polly audio quickly to callers, falls back to a secondary vendor when needed, and uses Hystrix to manage vendor failover. The Amazon Polly-based solution delivered higher-quality voices, better interpretation and streaming speed, and significant cost savings — Bandwidth reports a 78% cache hit rate, dramatically reducing the number of TTS requests.

Bandwidth

Yiannis Philipopoulos

Sr. Software Developer