Aniwaffle by Narya.ai

Build real-time Agents with a voice and a face.

Aniwaffle is Narya.ai’s platform for real-time voice Agents and Avatars. Every guide below starts from a fresh workspace and tells you exactly where to click.

Choose your path

What are you building?

Pick the closest outcome. You can move between products at any time.

Agents & Live

Build an AI Agent

Create a talking Agent, choose its voice and Avatar, then test the conversation live.

Start in the Dashboard. No code required.

Avatar

Bring an Avatar to life

Choose a character, preview it with speech, and publish it as part of an Agent.

Start visually. Rendering details come later.

Text to Speech

Generate speech

Use a ready-made voice or clone your own, then generate audio through the API.

Dashboard first, API when you need it.

Developers

Build with code

Embed a published Agent or integrate directly with the API. Credentials come from the Dashboard; the authentication guide shows every step.

SDK on npm

JavaScript client

Connection lifecycle, transcripts, text input, and push to talk.

@animated-waffle/client SDK on npm

React

The client API plus Agent-bound Avatar rendering and React state.

@animated-waffle/react SDK on npm

REST & WebSocket API

Direct server integration for TTS, Agents, sessions, and catalogs.

/v1

Platform boundary

Built for managed realtime

Server-safe by default

Permanent workspace keys stay on your backend. Browsers receive short-lived, scoped credentials.

Realtime is built in

Managed sessions handle speech, interruption, audio transport, and Avatar performance together.

One published identity

Voice, Avatar, prompt, and model come from the Agent publication instead of browser overrides.