Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.tavus.io/llms.txt

Use this file to discover all available pages before exploring further.

What is a replica?

A replica is a realistic AI avatar that can speak, listen, learn, and respond like a real person. With Tavus, you can create a replica that looks and behaves like you—or use a stock replica to generate conversations instantly. Replicas are designed to capture natural expressions, tone, and presence, making interactions feel more human and engaging.  Replicas are generated using our Phoenix-4 model, built on advanced research in real-time human rendering and trained on large-scale conversational data to model natural expression, emotion, and behavior. 

Custom Replicas

You can create a replica in one of two ways:
  • Upload a training video – Record a 1-minute video with 30 seconds of speaking followed by 30 seconds of listening. This method provides the highest quality and most control. With this short input, Tavus can reproduce appearance with high fidelity, voice, expressions, and accurate lip-sync. See Training from a Video.
  • Upload an image – Upload a photo of a person or character, and Tavus will automatically generate the training data and create the replica for you. This path is faster to set up but lower fidelity than video. See Training from an Image.
Creating a Custom Replica is only available on the Starter, Growth, and Enterprise plans.

Stock Replicas

Alternatively, Tavus offers 100+ ready-to-use stock replicas recorded using real actors - great for quick prototyping, demos, or teams without access to custom training footage or images.

Key Features

Realistic Face Cloning

Custom replicas capture a person’s look, expressions, and speaking style.

Emotional Control

Phoenix-4 replicas express emotions through lifelike facial expressions while speaking and listening.

Full Motion Control

Smooth transitions between speaking and listening with fully generated, natural head motion.
Emotional expression controls in conversational video apply only to Phoenix-4; see Emotion Control with Phoenix-4.

Platform Policies

By using the Tavus platform and APIs, you agree to comply with Tavus’ Terms of Service and Acceptable Use Policy. You are responsible for ensuring you have the necessary rights and permissions to upload and use any training data, including videos, images, voices, and likenesses.

Getting Started

You can create a replica using either the Developer Portal or the API, depending on your workflow.