Audio Privacy
Clients stream audio directly to STT providers; the backend handles credentials, state, logs, and transcripts only.
About
The architecture keeps church audio off SermonSync servers, lets operators override AI instantly, and prioritizes projection reliability over novelty.
Clients stream audio directly to STT providers; the backend handles credentials, state, logs, and transcripts only.
Consensus can auto-project, but disagreement, single-model detections, and paraphrases remain suggestions.
Every backend query and realtime room is scoped to a church/session boundary for tenant isolation.