Anam Avatar Plugin for Vision Agents

Add real-time interactive avatar video to your AI agents using Anam's avatar API.

Features

• Real-time avatar video and audio synchronized with TTS output
• Audio passthrough mode for high-quality voice reproduction
• Automatic interruption when the user starts speaking
• Works with any TTS provider (ElevenLabs, Cartesia, etc.) or Realtime LLMs
• Configurable video resolution (default 1920x1080)

Installation

uv add "vision-agents[anam]"
# or directly
uv add vision-agents-plugins-anam

Quick Start

import asyncio
from uuid import uuid4
from dotenv import load_dotenv

from vision_agents.core import User, Agent
from vision_agents.plugins import anam, deepgram, elevenlabs, getstream, gemini

load_dotenv()


async def start_avatar_agent():
    agent = Agent(
        edge=getstream.Edge(),
        agent_user=User(name="AI Assistant with Avatar", id="agent"),
        instructions="You're a friendly AI assistant.",

        llm=gemini.LLM(),
        tts=elevenlabs.TTS(),
        stt=deepgram.STT(),

        avatar=anam.Avatar(),
    )

    call = await agent.create_call("default", str(uuid4()))

    async with agent.join(call):
        await agent.simple_response("Say hello.")
        await agent.finish()


if __name__ == "__main__":
    asyncio.run(start_avatar_agent())

Configuration

Environment Variables

ANAM_API_KEY=your_anam_api_key
ANAM_AVATAR_ID=your_anam_avatar_id

Avatar Options

anam.Avatar(
    avatar_id=None,  # Anam avatar ID (or set ANAM_AVATAR_ID env var)
    api_key=None,  # Anam API key (or set ANAM_API_KEY env var)
    client_options=None,  # Optional Anam ClientOptions for advanced configuration
    connect_timeout=None,  # Seconds to wait for connection (None = wait indefinitely)
    session_ready_timeout=None,  # Seconds to wait for session ready (None = wait indefinitely)
    width=1920,  # Output video width in pixels
    height=1080,  # Output video height in pixels
)

How It Works

1. Anam Connection: Connects to Anam's avatar service and waits for the session to become ready
2. Audio Forwarding: TTS audio from the agent is resampled to 24kHz mono and streamed to Anam
3. Avatar Generation: Anam generates synchronized avatar video and lip-synced audio from the input
4. Video Streaming: Avatar video and audio frames are streamed back to call participants via GetStream Edge

Requirements

• Python 3.10+
• Anam API key and avatar ID (get them at anam.ai)
• GetStream account for video calls
• TTS provider (ElevenLabs, Cartesia, etc.) or Realtime LLM

License

MIT

Links

• Documentation
• GitHub
• Anam Docs