❗️
Interactive Avatar is getting upgraded to LiveAvatar! Now with its own website and app. Learn more
We recommend integrating directly with LiveAvatar instead. Docs
If you have an existing HeyGen Interactive Avatar implementation, we recommend checking out our migration guide. Docs

This endpoint is used to initiate a new streaming session with an Interactive Avatar. It sets up a fresh session, allowing for real-time interactions and communication.

Request Body

Field	Type	Description
`quality`	string	The quality of the data to be retrieved. Can be "high", "medium", or "low". high: 2000kbps and 720p. medium: 1000kbps and 480p. low: 500kbps and 360p.
`avatar_id`	string (optional)	The ID of the Interactive Avatar to use. If not provided, a default avatar will be chosen. Default: `default`
`voice`	VoiceSetting(optional)	The settings for the Interactive Avatar's voice.
`stt_settings`	STTSetting(optional) beta	The speech-to-text settings control how audio is converted into text.
`video_encoding`	string (optional)	Specifies the encoding format for streaming video. Can be "H264", "VP8". Default: `VP8`. Note: Choosing "H264" may offer better compatibility with certain devices and platforms.
`knowledge_base`	string (optional)	Knowledge Base prompt used for chat task type.
`version`	string (optional) beta	Specifies the version to use. Currently, the only valid value is `v2`. Default: Not specified (uses v1).
`knowledge_base_id`	string (optional) beta	The ID of the knowledge base to use for the avatar's responses. Only applicable when `version` is set to `v2`.
`disable_idle_timeout`	boolean(optional)	By default session has a 2 minute idle timeout, setting to true disables it. ⚠️ Do not use this feature without proper session management, as open sessions can consume your API credits!
`activity_idle_timeout`	integer(optional)	Specifies the maximum idle time in seconds allowed after the last user activity before the session is considered inactive. Default value: 120 seconds. Min value: 30 seconds, Max value: 3600 seconds.
`livekit_settings`	LiveKitSettings (optional) beta	Settings for connecting to your own LiveKit instance. ⚠️ Only provide this if you are using your own LiveKit, not HeyGen-managed LiveKit.

VoiceSetting

Field	Type	Description
`voice_id`	string (optional)	Voice for your Interactive Avatar. See the available voices by calling the List Voices endpoint. Note: Not every voice is supported in the streaming API.
`rate`	float(optional)	Voice speed rate. Default is 1.
`emotion`	string (optional)	Emotion to use for Emotional voices. Available emotions are `Excited`, `Serious`, `Friendly`, `Soothing`, `Broadcaster`
`elevenlabs_settings`	ElevenlabsSettings (optional)	Voice settings to pass over if the voice provider for the session is Elevenlabs.

ElevenlabsSettings

Field	Type	Description
`stability`	float (optional)	Default is 0.75.
`model_id`	string(optional)	Voice model id. For ElevenLabs available models is: `eleven_flash_v2_5`, `eleven_multilingual_v2`. Default: `eleven_flash_v2_5` Note: Some languages are known to be problematic with flash. Try multilingual_v2 if you're having problems.
`similarity_boost`	float(optional)	Default is 0.75.
`style`	float (optional)	Default is 0.0.
`use_speaker_boost`	bool (optional)	Default is true.
`speed`	float (optional)	Default is 1.0, min is 0.7, max is 1.2

STTSettings

Field	Type	Description
`provider`	string (optional) beta	STT model. Allowed values: `deepgram, gladia, assembly_ia`. `assembly_ai`is default provider for English language transcription.
`confidence`	float(optional)	Default is 0.55.

LiveKitSettings

Field	Type	Description
`room`	string (optional)	The LiveKit room to join. Must match the token’s `room` claim.
`url`	string (optional)	Your LiveKit server URL. Example: `wss://mylivekit.example.com`
`token`	string (optional)	A valid LiveKit access token with the required permissions.

Response

Field	Type	Description
`code`	integer	The response status code.
`message`	string	A message providing more details about the request's result, typically explaining success or error.
`data`	object	Contains the main data for the response.
`data.session_id`	string	A unique identifier for the session.
`data.url`	string	The WebSocket URL for accessing the LiveKit room (e.g., `wss://heygen-feapbkvq.livekit.cloud`).
`data.access_token`	string	The access token required to authenticate and join the LiveKit room.
`data.session_duration_limit`	integer	The maximum allowed duration (in seconds) for the session, indicating any session time limits.
`data.is_paid`	boolean	A flag indicating whether the session is part of a paid plan.
`data.realtime_endpoint`	string	The real-time endpoint URL for alpha implementations, which could be used for experimental features (e.g., `wss://webrtc-signaling.heygen.io/v2-alpha/...`).
`data.livekit_agent_token`	string	Only returned when using HeyGen-managed LiveKit. This token is for HeyGen’s audio agents (e.g., Pipecat / LiveKit Agent) to join the room and provide audio services. Not needed when using your own LiveKit instance.