Resolved. All voices are available again and the agent speaks normally in all conversations. The missing voices have been restored in the new EU environment, and we are no longer seeing failures since 15:11 CEST. Conversations that were affected between 11:44 and 15:11 CEST can simply be restarted. No customer data was affected.

Since today 09:44 UTC, a portion of conversations start but the agent stays silent – speech synthesis fails while transcription and responses work normally. The cause is identified: after a change to our speech provider's EU hosting, part of the voice catalog is not yet available in the new environment. Conversations using an affected voice end without spoken output.

We have identified the fix and are restoring the missing voices now. No customer data is affected. Next update within the hour.