Avatars

An avatar gives your voice agent a face: a real-time video presence that lip-syncs to the agent’s speech and publishes a video track into the room. Add one by passing an avatar to the pipeline’s avatar slot; it works with both Cascade and Realtime pipelines.

# anam_avatar = AnamAvatar(...) - see Avatar Integration below
from zrt.inference import AICousticsDenoise

pipeline = Pipeline(
    stt=DeepgramSTT(), llm=OpenAILLM(), tts=ElevenLabsTTS(),
    vad=SileroVAD(), turn_detector=TurnDetector(model="echo-large"),
    avatar=anam_avatar,   # ← the agent now has a visual face
    denoise=AICousticsDenoise(model_id="quail-vf-2.2-l-16khz"),
)

Integration

Use a managed avatar provider: Anam or Simli.

Options

Approach	When to use
Managed provider (Anam, Simli)	Drop-in photorealistic avatars with one API key.

Avatar providers (Anam and Simli) are available in the Python, JavaScript, and Go SDKs.

References

Python

SDK Reference

AgentSession

AgentSession in the Python API reference.

Video Input Avatar Integration

Get started

Modalities

Configure Pipeline

Tools and Capabilities

Turn Detection and Interruptions

Telephony Agents

Context Management

Observability and Monitoring

Integration

Options

References

SDK Reference

AgentSession

Integration

​Options

​References

​SDK Reference

AgentSession

Options

References

SDK Reference