LemonSlice released Character World Model-1 on September 23, introducing an interactive avatar model that generates a character’s body and surrounding scene as a conversation unfolds.
The company’s announcement says CWM-1 can start from an image and use audio and prompts to generate the video. An emotion-control system lets developers influence gestures and actions. A public demo is free to try, while API access is listed for Enterprise and Ultra customers.
The product goes beyond lip movement
LemonSlice describes a system that generates the scene as well as the face, including hand movements and interactions with objects. These are the company’s stated capabilities; OSBBD has not independently tested their consistency.
For a developer, visual quality during a short demonstration is only one criterion. Interruptions, changes of topic and long sessions can reveal behavior that a selected clip does not.
The product’s controls also need to fit the intended use. A character that looks expressive may still produce a distracting gesture or an inappropriate response for the application.
Realistic presentation raises a disclosure question
An avatar interface should make it clear that the user is interacting with generated media. Visual realism does not establish that a person is present or that the system understands the conversation.
Before integrating an image-based avatar, a team also needs an appropriate right to use the image and a clear account of how uploaded material is handled. Our AI upload guide separates processing, storage and training policies.
