[Suggestion] Zoom RTMS real-time meeting transcription with speaker diarization (Python)

[deepgram-robot]

What to build

A working example that connects to Zoom's Real-Time Media Streams (RTMS) API to capture live meeting audio and transcribes it in real time using Deepgram's streaming STT with speaker diarization — producing a live, speaker-labelled transcript during an active Zoom meeting.

Why this matters

Zoom RTMS is an emerging API that gives developers access to live meeting audio streams without screen sharing or recording. Developers building meeting intelligence products (real-time notes, compliance monitoring, coaching tools) need a reference implementation showing how to wire Zoom's RTMS WebSocket to Deepgram's streaming STT. This is a high-demand integration — millions of meetings happen daily on Zoom, and the RTMS API is the supported path for real-time access.

Suggested scope

• Language: Python
• Deepgram APIs: Streaming STT (Nova-3) with diarization, smart formatting
• Dependencies: @zoom/rtms SDK (or direct WebSocket), deepgram-sdk
• Architecture:
  • Zoom RTMS webhook receiver (meeting start/end events)
  • RTMS WebSocket client receiving raw audio
  • Deepgram streaming STT session with diarization enabled
  • Real-time transcript output with speaker labels and timestamps
• Includes: Zoom App configuration guide, ngrok setup for local development
• Complexity: Medium — requires Zoom App setup and two concurrent WebSocket connections

Acceptance criteria

• Runnable with minimal setup (clone, add API keys, configure Zoom App, run)
• README explains Zoom RTMS setup step by step
• Uses current SDK version
• Produces real-time speaker-diarized transcript during a live Zoom meeting
• Handles meeting join/leave events and connection lifecycle
• Includes sample output showing the transcript format

---

Raised by the DX intelligence system.