Skip to content

Conversation

@YehoavR
Copy link
Contributor

@YehoavR YehoavR commented Dec 31, 2025

Summary

  • Add avatar-live realtime model with set_image() method for dynamic avatar updates
  • Add lucy-restyle-v2v video model with reference_image support (mutually exclusive with prompt)
  • Add AvatarOptions dataclass for configuring avatar image on connect
  • Add VideoRestyleInput with validation for prompt/reference_image mutual exclusivity
  • Add initial_prompt option for sending prompt before WebRTC handshake
  • Update image ACK message format to use set_image_ack with success/error fields

Changes

Avatar Live (Realtime API)

  • New model: models.realtime("avatar-live")
  • New AvatarOptions for passing avatar image on connect
  • New InitialPromptOptions for sending prompt before WebRTC handshake
  • New set_image() method on RealtimeClient for runtime avatar updates
  • WebSocket message types: SetAvatarImageMessage, SetImageAckMessage
  • Recv-only video transceiver for avatar-live mode

Message Types

  • Updated SetImageAckMessage with success: bool and error: Optional[str]
  • Added ErrorMessage for server error handling
  • Added ReadyMessage for server ready signal
  • Added IceRestartMessage with TurnConfig for ICE restart support

Reference Image (Queue API)

  • New model: models.video("lucy-restyle-v2v")
  • Support for reference_image parameter (mutually exclusive with prompt)
  • Validation: enhance_prompt only valid with text prompt

Test plan

  • All 82 tests pass
  • Tests for lucy-restyle-v2v with prompt
  • Tests for lucy-restyle-v2v with reference_image
  • Tests for mutual exclusivity validation
  • Tests for avatar-live model definition
  • Tests for avatar-live connect with avatar image
  • Tests for set_image() method
  • Tests for set_image() only available on avatar-live
  • Tests for set_image() timeout handling
  • Tests for set_image() server error handling
  • Tests for initial_prompt option

- Add avatar-live realtime model with set_image() method
- Add lucy-restyle-v2v video model with reference_image support
- Add AvatarOptions for configuring avatar image on connect
- Add VideoRestyleInput with mutual exclusivity validation
- Add comprehensive tests for both features
- Introduced a Gradio-based test UI for easy feature testing.
- Added functionality for image and video generation, transformation, and restyling.
- Included a tokens API for creating short-lived client tokens.
- Updated README with instructions for using the test UI.
- Update image ACK message type from 'image_set' to 'set_image_ack'
- Change ACK format to use success:bool + error:string
- Add initial_prompt option for sending prompt before WebRTC handshake
- Add ErrorMessage, ReadyMessage, and IceRestartMessage types
- Add ICE restart with TURN server support
- Update tests for new message format and initial_prompt feature
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant