Avatar Integration - Zero Runtime

The quickest way to add a face is a managed avatar provider. Construct the avatar and pass it to the pipeline’s avatar slot; the SDK streams the agent’s speech to the provider, which renders a lip-synced video track into the room.

Anam

Set your Anam credentials, then construct AnamAvatar:

export ANAM_API_KEY=<key>

from zeroruntime.plugins import AnamAvatar
from zeroruntime.inference import AICousticsDenoise

avatar = AnamAvatar(
    avatar_id="<avatar-id>",   # falls back to ANAM_AVATAR_ID / a default
)

pipeline = Pipeline(
    stt=DeepgramSTT(), llm=OpenAILLM(), tts=ElevenLabsTTS(),
    vad=SileroVAD(), turn_detector=TurnDetector(model="echo-large"),
    avatar=avatar,
    denoise=AICousticsDenoise(model_id="quail-vf-2.2-l-16khz"),
)

import { AnamAvatar } from '@zeroruntime/js-sdk/plugins';
import { AICousticsDenoise } from '@zeroruntime/js-sdk/inference';

const avatar = AnamAvatar({
  avatar_id: '<avatar-id>',  // falls back to ANAM_AVATAR_ID / a default
});

const pipeline = Pipeline({
  stt: DeepgramSTT(), llm: OpenAILLM(), tts: ElevenLabsTTS(),
  vad: SileroVAD(), turn_detector: TurnDetector({ model: 'echo-large' }),
  avatar,
  denoise: AICousticsDenoise({ model_id: 'quail-vf-2.2-l-16khz' }),
});

Parameter	Type	Default	Description
`api_key`	`str`	`None`	Anam API key. Falls back to `ANAM_API_KEY`.
`avatar_id`	`str`	a default	The avatar to render. Falls back to `ANAM_AVATAR_ID`.
`persona_name`	`str`	`None`	Optional persona name for the avatar.
`voice_id`	`str`	`None`	Optional voice override for the avatar.

Simli

Set your Simli credentials, then construct SimliAvatar with the face to render:

export SIMLI_API_KEY=<key>

from zeroruntime.plugins import SimliAvatar
from zeroruntime.inference import AICousticsDenoise

avatar = SimliAvatar(
    face_id="<face-id>",   # falls back to SIMLI_FACE_ID / a default
)

pipeline = Pipeline(
    stt=DeepgramSTT(), llm=OpenAILLM(), tts=ElevenLabsTTS(),
    vad=SileroVAD(), turn_detector=TurnDetector(model="echo-large"),
    avatar=avatar,
    denoise=AICousticsDenoise(model_id="quail-vf-2.2-l-16khz"),
)

import { SimliAvatar } from '@zeroruntime/js-sdk/plugins';
import { AICousticsDenoise } from '@zeroruntime/js-sdk/inference';

const avatar = SimliAvatar({
  config: { faceId: '<face-id>' },  // Simli's own config object carries the face
});

const pipeline = Pipeline({
  stt: DeepgramSTT(), llm: OpenAILLM(), tts: ElevenLabsTTS(),
  vad: SileroVAD(), turn_detector: TurnDetector({ model: 'echo-large' }),
  avatar,
  denoise: AICousticsDenoise({ model_id: 'quail-vf-2.2-l-16khz' }),
});

Parameter	Type	Default	Description
`api_key`	`str`	`None`	Simli API key. Falls back to `SIMLI_API_KEY`.
`face_id`	`str`	a default	The face to render. Falls back to `SIMLI_FACE_ID`.
`model`	`str`	`None`	Optional Simli model override.
`is_trinity`	`bool`	`False`	Set when the face ID is a Trinity avatar, to keep sync.