Headless AI avatar API, made in Japan

Real-time AI avatars that speak many languages

avacast is a headless AI avatar API made in Japan. Add an AI avatar that holds real-time conversations to websites, apps, reception kiosks, digital signage and more. It speaks Japanese, English, Chinese and many other languages.

avacast screen

Made in Japan

Developed and operated by SHANNON. Contracts and support are in Japanese, with a Japanese company.

Headless

avacast streams the avatar’s video and audio; you decide the screen design and layout.

Multilingual

Japanese, English, Chinese and more: the avatar speaks in the language of the text.

Demo

How it works

Actual streamed footage. Lips sync to the audio and the head turns naturally.

Build it from an AI agent (MCP):https://api.avacast.jp/v1/mcp

Features

What avacast does

Developed and operated by SHANNON

SHANNON handles avacast from development through operation. Dedicated GPU nodes are also available.

  • Voices selectable by age and gender
  • Original avatars from a single photo
  • Dedicated GPU nodes available
An avatar giving directions on shopping mall signage

A headless design that fits your use case

avacast streams the avatar’s video and audio. You decide the screen design and layout for your use case, and it runs anywhere a browser does: websites, apps, reception kiosks and signage. What the avatar says can be a prepared script or answers from your chatbot or LLM.

  • Four APIs: sessions, speak, interrupt and close
  • Browser SDK loaded with a single script tag
  • MCP support for integration by AI agents such as Claude
avacast API docs

Multilingual, including Japanese, English and Chinese

The avatar speaks in the language of the text you send. Lip movement is generated from the audio, so the mouth matches naturally in every language.

  • Guidance for overseas visitors and help desks for international staff
  • One avatar for multiple languages
  • Speaks in the language of each text
avacast console: avatars
Flow

Getting started

From sign-up to production.

  1. Sign up

    Sign up with Google or email and issue an API key in the console.

  2. Create a session

    Your server creates a session with the API key and receives a short-lived token for the browser.

  3. Integrate

    Pass the token to the SDK, and video and audio stream over WebRTC. Send what the avatar should say as text.

  4. Go live

    Check connected minutes, caps and live sessions in the console.

FAQ

FAQ

Use a prepared script, or pass your chatbot’s or LLM’s answers as they are. avacast turns the text into video of the avatar speaking.
Japanese, English, Chinese and many other languages. The avatar speaks in the language of the text.
Yes. An avatar can be created from a single photo in a few minutes. Use a bright, front-facing photo with the mouth closed.
Pricing and plan details are on the pricing page of the product site. Sign up with an email address or a Google account.

Talk to us about avacast.

We reply within two business days.