Live Meeting Transcription with ElevenLabs Scribe v2 Realtime

LLazare Rossillon

Stream live transcripts out of Zoom, Google Meet and Microsoft Teams with ElevenLabs Scribe v2 Realtime (~150ms latency): a bot joins the call, your WebSocket receives text as people speak.

Live Meeting Transcription with ElevenLabs Scribe v2 Realtime
August 13, 2026

Post-call transcripts are table stakes. The interesting products — live captions, in-meeting copilots, agents that answer during the call — need text while people are speaking. ElevenLabs' Scribe v2 Realtime does streaming speech-to-text at roughly 150ms latency with diarization for up to 32 speakers across 90+ languages.

The missing half is live meeting audio. A meeting bot with streaming_config provides it: the bot sits in the Zoom, Google Meet or Microsoft Teams call and streams to your endpoint in real time, with ElevenLabs as the streaming transcription engine.

ElevenLabs is the streaming-only provider in the Meeting BaaS lineup: it's selected for real-time transcription, while post-call transcripts use Gladia or your own batch provider.

Step 1: Get your keys

  • Meeting BaaS: self-serve signup at auth.meetingbaas.com — first 8 recording hours free.
  • ElevenLabs: an API key from your ElevenLabs account if you're bringing your own.

Step 2: Send a bot with live transcription

curl -X POST "https://api.meetingbaas.com/v2/bots" \
     -H "Content-Type: application/json" \
     -H "x-meeting-baas-api-key: $API_KEY" \
     -d '{
           "meeting_url": "https://us02web.zoom.us/j/1234567890",
           "bot_name": "Live Captions Bot",
           "streaming_config": {
             "mode": "transcription",
             "output_url": "wss://your-app.example.com/live-transcript",
             "audio_frequency": 24000
           },
           "transcription_config": {
             "provider": "elevenlabs",
             "api_key": "'"$ELEVENLABS_API_KEY"'"
           }
         }'

Your WebSocket at output_url receives the live transcript stream while the meeting runs. Prefer raw audio to run your own streaming model? Set mode to "audio" and you get the meeting's audio at 16, 24, 32 or 48 kHz instead.

Step 3: Build the live experience

With text arriving mid-meeting you can:

  • Live captions rendered in your product while the call happens
  • In-meeting copilots that surface answers, docs or CRM context as topics come up
  • Real-time compliance flags on regulated phrases, minutes before the call ends
  • Speaking agents: stream audio back through input_url and the bot talks — the full loop for interactive meeting agents

The standard pipeline still runs alongside: full recording, post-call transcript and metadata by webhook after the meeting.

What this costs

Streaming adds 0.10 tokens per hour per direction on top of the 1 token per hour recording rate — a live-transcribed, recorded meeting hour stays under 1.5 tokens ($0.53–0.75 depending on token pack). ElevenLabs usage on your own key is billed by ElevenLabs directly.

Next steps

Similar blogstutorial