Context-Aware Motion Generation Technology Based on Text and Emotion Embeddings

Context-Aware Motion Generation Technology Using Text and Emotion Embeddings

This technology generates motion sequences with emotional context by vectorizing text prompts and emotional information into separate embeddings, then applying the emotion embedding as a condition within a generative AI model.

Existing text-based motion generation often struggles to capture the nuances or intensity of emotions, making it difficult to differentiate the same action based on the underlying emotional state.

By controlling the diffusion process with emotion embeddings and weights, and adjusting facial, hand, and body movements accordingly, this technology enables the creation of expressive, context-aware motions for 3D character content.

‍

‍

Key Features:

• A processor and memory that receive input information, including text prompts for character actions and the specific emotions to be conveyed.
• Text embeddings generated from prompts and emotion embeddings that determine the nuance of the motion.
• A generative AI model that creates motion sequences based on text embeddings while incorporating emotion embeddings as a condition to provide emotional context.
• Emotion weights that adjust based on the intensity of the emotional information, applied to the diffusion process or attention mechanism to vary the depth of emotional expression.

‍

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 further details.

‍