Personalized 3D Character Motion Generation Technology Based on User Attribute Metadata

Personalized 3D Character Motion Generation Technology Using User Attribute Metadata

This technology generates a user attribute context vector from text prompts and user metadata such as gender, age, and region. It then inputs this vector, along with text and biomechanical vectors, into a diffusion-based motion generation model to produce personalized joint rotation sequences.

Existing 3D character motion generation methods often fail to reflect the gender, age, or regional characteristics of the user or character, making it difficult to create motions tailored to individual traits.

By incorporating motion styles specific to each attribute through user attribute context vectors and contrastive learning, this technology can be applied to personalized 3D character content to generate motions that align with specific user characteristics.

‍

‍

Key Features:

• Acquiring user attribute metadata, including text prompts, gender, age, and regional codes, from a user terminal.
• Inputting the user attribute metadata into a user attribute encoder to generate a context vector containing values related to gender, age, and regional characteristics.
• Inputting the text condition vector, user attribute context vector, and biomechanical context vector into a diffusion-based motion generation model to produce joint-specific rotation sequences.
• Updating the user attribute encoder and model parameters through contrastive learning using motion data pairs with identical gender, age, and regional combinations as positive comparison pairs.

‍

This patent portfolio consists of four family patents covering AI-based 3D motion generation technology developed by Bumblebee Co., Ltd. Technology transfer is available alongside the patents, and the sale price is negotiable. Please refer to the attached technical manual for detailed information.

‍