Skip to main content

ADR-006: Voice (STT/TTS) via ElevenLabs

  • Status: Accepted
  • Date: 2026-02-04

Context​

We want voice tutoring as a premium-quality feature, with predictable vendor behavior and server-side control.

Decision​

  • Use ElevenLabs for STT and TTS.
  • Implement voice server-side only.
  • Do not offer voice to anonymous users.

Consequences​

  • We must broker audio streaming through the backend.
  • Quota and abuse controls must cover voice usage.

Alternatives considered​

  • Client-side direct-to-vendor: less control and harder to enforce quotas.