WebRTC Voice Test

Talk to the AI voice agent in the browser over real WebRTC

Ready

Bot server URL

Language

Use Case

Conversation Flow

Loopback mode

Echo mic back — no server needed

Click the button to start a test call

http://localhost:8765

Transcript will appear here during the call

1. Browser captures your mic and opens a real WebRTC peer connection.

2. SDP offer is POSTed to {botUrl}/api/voice/offer; the bot returns an answer.

3. The agent's voice streams back over WebRTC; transcripts arrive on /ws/voice/{id}.

• Speech detection, turn-taking and barge-in are handled server-side by pipecat (Sarvam native VAD).

• Just talk naturally — interrupt the agent any time.

• Requires the bot running with a valid .env (Sarvam + OpenAI keys).