All systems operational.
Static health for the hosted service, the latency budget we hold ourselves to, and recent incident history.
Conversation Loop
Median reply under a second
Speech Recognition
Streaming, committed on a silence threshold
Avatar Rendering
30 to 60 fps sustained
Meeting Studio
Recording and live transcripts
Avatar Editor
Live 3D preview
API and Webhooks
GraphQL and webhook delivery
790 ms serial
The stages overlap. A partial transcript is sent speculatively, and the first clause of a response is synthesized while the rest is still generating.
Capture and Voice Detection
40 ms
Speech to Text
180 ms
First Model Token
220 ms
First Audio Chunk
190 ms
First Rendered Frame
160 ms
- August 22, 2026Resolved
Recording export queue backlog
Server-side composition completed normally, but export delivery ran behind for a period. Live calls, transcripts and the conversation loop were unaffected.
- May 30, 2026Resolved
Avatar Editor preview failing to load
A shader compilation regression prevented the live 3D preview from initializing on a WebGL2 context with no float texture support. Resolved within the hour, and no calls were affected.