Skip to main content

​
Casting Director configuration (p735435f8c36)

{
  "pal_name": "Casting Director Julian",
  "pipeline_mode": "full",
  "default_face_id": "r92debe21318",
  "layers": {
    "perception": {
      "perception_model": "raven-1",
      "visual_awareness_queries": [
        "Is a second person clearly visible in the actor's frame, in addition to the actor?",
        "What is the actor's overall on-camera wardrobe style and color, the kind of styling that could read as a costume choice for a role?"
      ],
      "audio_awareness_queries": [
        "Does the actor sound nervous or confident?",
        "Does the actor's vocal energy rise or fall during the read?"
      ],
      "perception_analysis_queries": [
        "Rate the actor's overall camera presence, energy, and range during the audition.",
        "Did the actor's energy and engagement increase or decrease over the course of the audition?",
        "Was the actor alone for the entire audition?",
        "Was there any indication of external coaching, another voice, whispering, or the actor receiving prompts off-screen?",
        "Aside from the sides shown on screen during the cold read, did the actor appear to be reading from hidden notes or a second screen off to the side?"
      ]
    },
    "tts": {
      "tts_emotion_control": true,
      "tts_model_name": "sonic-3"
    },
    "llm": {
      "model": "tavus-glm-4.7",
      "speculative_inference": true
    },
    "conversational_flow": {
      "turn_detection_model": "sparrow-1",
      "turn_taking_patience": "high",
      "voice_isolation": "near",
      "idle_engagement": "off",
      "pal_interruptibility": "low"
    }
  }
}
This template PAL (Casting Director Julian) includes:
  • PAL identity: A casting director for interactive auditions and table-reads: objectives, guardrails, and Raven perception for on-camera presence and read quality.
  • Full pipeline mode: Enables the full Tavus conversational pipeline, including perception, STT, LLM, and TTS.
  • System prompt: Omitted here for length. Retrieve the live template with Get PAL or clone it from PAL Maker.
  • Model layers:
    • Perception: Uses the raven-1 perception model.
    • TTS: sonic-3 with emotion control enabled.
    • LLM: tavus-glm-4.7 with speculative inference.
    • Conversational flow: sparrow-1 with high turn-taking patience, near voice isolation, and low interruptibility.

​
Create a conversation with this template PAL

1

Step 1: Create your conversation

Create a conversation using the template pal_id. When the PAL has a default_face_id, you can omit face_id:
cURL
curl --request POST \
  --url https://tavusapi.com/v2/conversations \
  --header 'Content-Type: application/json' \
  --header 'x-api-key: <api_key>' \
  --data '{
  "pal_id": "p735435f8c36"
}'
Replace <api_key> with your actual API key. You can generate one in the PAL Maker.
2

Step 2: Join the conversation

Click the link in the conversation_url field to join the conversation:
{
  "conversation_id": "c7f3fc6d788f",
  "conversation_name": "New Conversation",
  "conversation_url": "<conversation_link>",
  "status": "active",
  "callback_url": "",
  "created_at": "2025-05-20T05:38:51.501467Z"
}