Skip to Content
Introduction

Introduction

AvatarGo gives your app a talking avatar made from one photo. Use it two ways:

  • Live calls — the avatar talks with your user in real time, in your web page: it listens, answers by voice with its lips in sync, shows captions, and stops when the user cuts in.
  • Pre-render — a video of the avatar saying a script you write, to download and use anywhere: on your site, in a course, in a message.

Live calls

A call is with a character: a face, a voice, and what it knows and how it behaves — its instructions, a greeting, documents it can look things up in, and tools it can use, such as a lookup on your server or an action in your page. You build characters in the AvatarGo console  and try them there with a test call. See Characters.

To put a call in your page:

  1. Your server creates a session with your API key and passes the page a session token for that call.
  2. Your page joins the session with the JavaScript SDK and shows the avatar in a <video> element.

Your API key stays on your server; the page only ever holds the session token. The Quick start does both steps in a few lines.

Already have your own speech recognition, language model and voice? Use avatar only: your server sends the audio and the avatar lip-syncs to it — no character needed. See Avatar only (your own audio).

Pre-render

Give a picture, a voice and a script, and get back a video (MP4) of the avatar saying it, with subtitles (SRT) if you want them. Two kinds:

  • Talking head — head and shoulders, from a photo of a face.
  • Half-body — the whole picture kept as it is, hands and props included; it also works with an illustration. It takes longer to make.

Make one on the console’s Pre-render page, or from your server with POST /v1/renders. A script is up to 4,000 characters, and a video up to 5 minutes. A video is made in the background: you see it once it is ready, and it is kept until you delete it. See Pre-render.

Where to go next

  • Quick start — your first call in a web page.
  • Reference — every function and option of the JavaScript SDK.
  • Events — captions, tool calls and how a call ends, as they happen.
Characters