Echo Mode gives you full control over the Replica’s speech by letting you bypass Tavus’s core layers and supply your own input directly. There are two echo modes:

Echo Mode is incompatible with perception. As a result, we recommend using Tavus’s Full Pipeline in its entirety for the lowest latency and most optimized multimodal experience. Integrations like LiveKit Agent or Pipecat only provide rendering, while our Full Pipeline includes perception, turn-taking, and rendering for complete conversational intelligence.

​
Text or Audio (Base64) Echo

Microphone is disabled in the Transport layer.

  • Text Echo

    • Bypasses Perception, STT, and LLM
    • Sends raw text directly to the TTS Layer
    • Useful for manually scripted speech and interrupt control

  • Audio (Base64) Echo

    • Bypasses all layers except the Realtime Replica Layer
    • Sends base64-encoded audio for direct playback by the Replica

Send text or base64 audio using the Interactions Protocol.

​
Microphone Echo

Microphone is enabled in the Transport layer.

  • Bypasses all CVI layers (Perception, STT, LLM, TTS)
  • Streams pre-generated audio directly into the Replica
  • All interrupt logic must be embedded in your audio stream

​
Echo Mode Quickstart

​
Prerequisites

Before you begin, ensure the following dependencies are installed:

​
Create an Echo Mode Conversation

1

Step 1: Get an API Key

  1. Go to the Tavus platform and select API Key from the sidebar menu.
  2. Click Create New Key to begin generating your API key.
  3. Enter a name for the key and (optional) specify allowed IP addresses, then click Create API Key.
  4. Copy your newly created API key and store it securely.

We cannot recover your API Key if you lose it.

2

Step 2: Create an Echo Persona

Use the following request to create a persona:

cURL
curl --request POST \
  --url https://tavusapi.com/v2/personas \
  --header 'Content-Type: application/json' \
  --header 'x-api-key: <api-key>' \
  --data '{
    "persona_name": "Echo Assistant",
    "pipeline_mode": "echo"
  }'
3

Step 3: Create a Conversation

Create a conversation with your newly created persona_id:

cURL
curl --request POST \
  --url https://tavusapi.com/v2/conversations \
  --header 'Content-Type: application/json' \
  --header 'x-api-key: <api_key>' \
  --data '{
    "persona_id": "<your_persona_id>",
    "conversation_name": "Echo Test"
  }'
4

Step 4: Create an App

Create a file named script.py and paste the following code:

import sys
from flask import Flask, jsonify, request
from daily import CallClient, Daily, EventHandler

app = Flask(__name__)

# Global client instance
call_client = None

class RoomHandler(EventHandler):
    def __init__(self):
        super().__init__()

    def on_app_message(self, message, sender: str) -> None:
        print(f"Incoming app message from {sender}: {message}")

def join_room(url):
    global call_client
    try:
        Daily.init()
        handler = RoomHandler()
        call_client = CallClient(event_handler=handler)
        call_client.join(url)
        print(f"Joined room: {url}")
    except Exception as e:
        print(f"Error joining room: {e}")
        raise

@app.route("/send_text_message", methods=["POST"])
def send_text_message():
    global call_client
    if not call_client:
        return jsonify({"error": "Not connected to a room"}), 400

    try:
        body = request.json
        conversation_id = body.get("conversation_id")
        properties = body.get("properties", {})
        message = {
            "message_type": "conversation",
            "event_type": "conversation.echo",
            "conversation_id": conversation_id,
            "properties": {
                "modality": properties.get("modality"),
                "text": properties.get("text"),
                "audio": properties.get("audio"),
                "sample_rate": properties.get("sample_rate", 16000),
                "inference_id": properties.get("inference_id"),
                "done": properties.get("done")
            }
        }
        call_client.send_app_message(message)
        return jsonify({"status": "Message sent successfully"}), 200
    except Exception as e:
        return jsonify({"error": f"Failed to send message: {str(e)}"}), 500

if __name__ == "__main__":
    if len(sys.argv) != 2:
        print("Usage: python script.py <conversation_url>")
        sys.exit(1)

    conversation_url = sys.argv[1]
    try:
        join_room(conversation_url)
        app.run(port=8000, debug=True)
    except Exception as e:
        print(f"Failed to start the application: {e}")
        sys.exit(1)

This script starts a Flask app that connects to a Daily room and sends echo interaction messages to your persona.

5

Step 5: Execute the Code

Run the script:

python script.py <conversation_url>

Replace <conversation_url> with the URL returned when creating your conversation.

6

Step 6: Send Echo Message

Send a POST request to your local server to trigger an echo message:

cURL
curl -X POST http://localhost:8000/send_text_message \
  -H "Content-Type: application/json" \
  -d '{
    "message_type": "conversation",
    "event_type": "conversation.echo",
    "conversation_id": "<conversation-id>",
    "properties": {
      "modality": "text",
      "text": "Hello there!",
      "audio": "base64-encoded-audio",
      "sample_rate": 24000,
      "inference_id": "inference-id-123",
      "done": "true"
    }
  }'

In this example, the persona will respond by saying: “Hello there!”