How I got a decent voice into a kids' app without breaking the bank or making anyone dig into settings

▤ Summary

This article explains how to build a reliable, low-cost text-to-speech system for a children’s app. It covers the move from a simple MVP using expo-speech to a more robust architecture using ElevenLabs, caching, hashing, and bundled audio fallbacks. It also describes how the setup avoids expensive API calls, keeps voices deterministic, and supports offline behavior. The piece includes implementation details for CI auditing, parity checks, and audio versioning to keep the voice experience consistent across platforms.

▦ Linked Companies