Azure TTS Startup Error

Hello Synthiam team,

Since today, both of my PiB robots (each running ARC on a separate Windows PC) have stopped speaking via the Azure Text To Speech Robot Skill. Their ARC startup sequences now fail.

The error reported by ARC is: "Server was unable to process request. ---> The type initializer for 'Microsoft.CognitiveServices.Speech.SpeechConfig' threw an exception. ---> Unable to find an entry point named 'register_send_callback' in DLL 'Microsoft.CognitiveServices.Speech.core.dll'."

My findings so far:

• There was no ARC update between the working and failing states, but Windows updates were installed.
• On both inspected PCs, ARC is version 2026.08.04.00. Windows update KB5129195 was installed on September 15.
• Windows audio and local Windows speech synthesis tests work through the robot’s speaker.
• The external Python programs run when launched directly from PyCharm.
• The ARC startup script calls the Azure TTS robot skill before launching the external Python programs. The speech exception appears to interrupt startup.
• The projects use German, French, and Greek Azure voices. A local Windows voice is not an equivalent replacement for these multilingual Azure voices.

Could you please clarify:

1. Does this error originate in Synthiam’s server-side Azure integration, Microsoft Azure, or the local ARC installation / Speech SDK on the PC?
2. Are you aware of a current incident or of a Speech SDK / native DLL version mismatch that could cause this?
3. Is there a fix or workaround that preserves multilingual Azure speech?
4. Is rolling back the Windows update (KB5129195) likely relevant to this specific error?

I have retained the original project and created a separate TEMP copy for controlled troubleshooting. I have not rolled back Windows updates.

Thank you for your help.