Developed and integrated real-time transcription with speaker diarization into the GetStream/Vision-Agents repository, enhancing its voice processing and transcription capabilities. The work centered on connecting the Mistral Voxtral API to enable accurate, multi-speaker audio transcription, with a focus on reducing latency and supporting complex audio workflows. Updated project documentation in Markdown to provide clear usage guidelines and reflect the new integration. This feature established a technical foundation for future multi-speaker transcription workflows, emphasizing real-time processing and robust API integration. The contribution demonstrates depth in handling real-time audio data and improving time-to-insight for users working with conversational audio content.