Complete documentation with detailed explanations and advanced configurations using the Livekit Agents Repository. This is helpful to see the various settings under the hood that are leveraged in the hedra-avatar-starter.
⚠️ Before you begin: Please make sure you have completed the Get Started section first to set up all required API keys and environment variables.

​
Hedra Avatar Setup

​
1. Using Avatar ID

​
Web UI

  1. Use Hedra Studio image generator to create an avatar image.
  2. Click to download your image.
  3. The UUID file name is the image asset id which can be used as avatar_id.

​
Hedra API

  1. Use the Hedra public api to upload an image asset to the platform.
  2. The id in the JSON response is the image asset id which can be used as avatar_id.

​
Add to .env

HEDRA_AVATAR_ID="<avatar_id>"

​
2. Image upload

Alternatively, you may pass an image from your local file system directly using the Livekit Hedra plugin:
  • You will still need a Hedra account and API key.
  • The image will be uploaded to Hedra’s platform where you will be able to view it in your library and get the id using one of the methods detailed above.

​
Installing Dependencies

​
1. Install the Livekit Hedra plugin

pip install "livekit-agents[hedra]~=1.0"
Note: You might need to use a virtual environment when installing dependencies using pip: 
python3 -m venv venv
source venv/bin/activate

​
2. Install other dependencies

As detailed in the Livekit docs. The included example uses OpenAI’s realtime model which can be installed with: 
pip install "livekit-agents[openai]~=1.0"

​
Avatar Example Setup

  1. Navigate to agents/examples/avatar_agents/hedra.
  2. Move your .env file with the necessary environment variables into this directory.
.env example: 
# Livekit
LIVEKIT_URL="wss://<project_name>.livekit.cloud"
LIVEKIT_API_KEY="<livekit_api_key>"
LIVEKIT_API_SECRET="<livekit_api_secret>"

# Hedra
HEDRA_AVATAR_ID="<avatar_image_id>" # omit if uploading image from local
HEDRA_API_KEY="<hedra_api_key>"

# Other providers
OPENAI_API_KEY="<openai_api_key>"
  1. If you choose to use a local file for your avatar image, you will need to edit the agent_worker.py example to pass the image properly:
from PIL import Image
avatar_image_path = "<path_to_local_image>"
avatar_image = Image.open(avatar_image_path)
hedra_avatar = hedra.AvatarSession(avatar_image=avatar_image)
await hedra_avatar.start(session, room=ctx.room)

​
Avatar Example Run

  1. From within the agents/examples/avatar_agents/hedra directory, run the following to start your agent in production mode:
python agent_worker.py start
Note: You may need to install missing dependencies with pip. Consider using a virtual environment: 
python3 -m venv venv
source venv/bin/activate

# Example
pip install dotenv
  1. Navigate to the Livekit Agents Playground in a browser.
  2. Connect to the project room that corresponds to the environment variables set in .env.
Your session request will not be queued until after a user connects to the room. It may take several seconds for your avatar to load.

​
Ending Your Avatar Session

​
1. Inactivity timeout

If the room does not get any user audio input for more than 2 minutes, the session will terminate due to user inactivity.
NOTE: this is currently disabled

​
2. Room disconnect

Click disconnect in the Livekit Playground and your session will stop.
NOTE: if you disconnect this way, you will incur an additional 20-30 seconds of billing charges because the session is designed to prevent immediate termination in the event of disconnects that may occur due to network instability

​
3. End your example process

End the agent_worker.py process using your favorite SIGINT, SIGKILL, etc. mechanism.

​
4. Use the Livekit CLI

If you’ve already set up the Livekit CLI (not covered in this guide) then you can run the following to terminate the room and session: 
lk room list
lk room delete <room_name>