Skip to main content

​
Getting Started with Tavus APIs

Tavus APIs let you build a Conversational Video Interface (CVI): real-time video conversations with a PAL. Each session pairs a PAL (behavior, knowledge, pipeline) with a face (photorealistic visual, Phoenix).
Terminology update: Tavus now uses PAL (behavior, knowledge, and pipeline configuration) and Face (visual appearance and voice) in the API and docs. Legacy names persona and replica still work on existing endpoints and request fields (/v2/personas, /v2/replicas, persona_id, replica_id, and related aliases) for backward compatibility.
You can access the API through standard HTTP requests, making it easy to integrate CVI into any application or platform.
For machine-readable API and docs context, use https://docs.tavus.io/openapi.yaml as the canonical HTTP API contract, https://docs.tavus.io/llms.txt as the docs page index, and https://docs.tavus.io/llms-full.txt as the full bundled docs export.

​
Who Is This For?

This API is for developers looking to add real-time, human-like AI interactions into their apps or services.

​
What Can You Do?

Use the end-to-end CVI pipeline to build human-like, real-time multimodal video conversations with these three core components:

PAL

Define the PAL’s behavior, tone, and knowledge.

Face

Train a lifelike digital face from a short video or headshot image.

Conversation

Create a real-time video call session with your PAL.
A typical path: Authentication, then Create PAL (POST /v2/pals), Create Face (POST /v2/faces), and Create Conversation. For what CVI includes end-to-end, see What is CVI?.