Semantic Avatar Rig for Diverse Facial Features

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing human avatar systems rely on generic base rigs that struggle to accurately represent diverse human facial features, particularly for under-represented demographics, requiring costly and time-consuming retrofitting to achieve fair representation.

Innovation Solution

A computing system that implements a semantic stylization avatar rig using identity parameters to generate a user face mesh, removing expression and pose parameters to create a neutral expression mesh, and applying semantic stylization techniques to incorporate diverse facial features such as hooded eyes, upturned noses, and skin tone, enhancing stylistic consistency through curvature analysis and image segmentation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of manufacture

If a generic base rig is used for avatar generation, then the system is simple and cost-effective, but it cannot accurately represent diverse human facial features particularly for under-represented demographics

Engineering Contradiction:
Improvecost-effectivenessVSAvoidaccuracy of diverse facial features
Core Design Contradiction:
Ease of manufactureVSManufacturing precision

Solution Approach 1:

The patent applies local quality by creating specific modular components for diverse facial features (hooded eyes, upturned noses, wide noses, thick lips, skin tones, hair types, accessories) that can be selectively applied to the generic base rig. This allows the system to maintain simplicity while adding localized diversity where needed, rather than redesigning the entire rig system.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The avatar generation system is segmented into distinct modular components: base rig, facial feature libraries, and styling elements. This segmentation allows diverse facial features to be added as separate, interchangeable modules to the generic base rig, enabling accurate representation of under-represented demographics without complicating the core system architecture.

Inventive Principle:
Principle #1Segmentation

2Manufacturing precision

If extensive asset libraries and complex interfaces are used to achieve fair representation of diverse demographics, then accuracy of diverse facial features is improved, but device complexity and cost increase

Engineering Contradiction:
Improveaccuracy of diverse facial featuresVSAvoidcomplexity of asset libraries and interfaces
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The generic base rig serves as a universal foundation that can accommodate multiple diverse facial features through modular additions. The same base rig structure works for all avatar types, and diverse features are added through universal interfaces that accept standardized modular components, reducing the need for multiple specialized systems.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent uses templated modular components that can be copied and applied to different avatars. Pre-defined facial feature modules (hooded eyes, upturned noses, etc.) are replicated and combined with the base rig through automated processes, reducing interface complexity while maintaining diverse representation accuracy.

Inventive Principle:
Principle #26Copying

3Manufacturing precision

If retrofitting is performed on existing generic rigs to achieve fair representation, then diverse facial features can be added, but it is time-consuming and costly

Engineering Contradiction:
Improverepresentation of diverse demographicsVSAvoidtime for retrofitting
Core Design Contradiction:
Manufacturing precisionVSLoss of time

Solution Approach 1:

Diverse facial feature modules are pre-prepared and organized in libraries before avatar generation. The system performs preliminary organization of hooded eyes, upturned noses, wide noses, thick lips, skin tones, hair types, and accessories into ready-to-use modules, eliminating the need for time-consuming retrofitting during actual avatar creation.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system enables self-service avatar generation where users can automatically assemble diverse facial features onto the base rig through simplified interfaces. The modular architecture allows the system to self-configure avatars by automatically matching and applying appropriate facial feature modules based on user selections, reducing manual retrofitting time.

Inventive Principle:
Principle #25Self-service

Data Source

PatentEP4600910A1Generation of semantically stylized avatars
Publication Date: 2025.08.13 MICROSOFT TECHNOLOGY LICENSING LLC
  • EP4600910A1 patent drawingFigure 1
  • EP4600910A1 patent drawingFigure 2~3B
  • EP4600910A1 patent drawingFigure 4A~4D

AI summary

A computing system (10) for generating semantically stylized avatars includes processing circuitry (18) that implements a semantic stylization avatar rig (26). In an inference phase, the processing circuitry (18) receives an instruction (32) to generate a semantically stylized avatar (34), the instruction (32) including an input image (36) of a user for which the semantically stylized avatar (34) is to be generated. A user face mesh (40) is generated based on identity parameters (42) of the input image (36), the user face mesh (40) including curves that determine a face size and a face shape, and relative proportions and positions of facial features. A neutral expression user face mesh (48) is generated by removing expression and pose parameters from the user face mesh (40). The input image (36), the identity parameters (42), and the neutral expression user face mesh (48) are input to a semantic stylization avatar generation module (50) to generate a semantically stylized avatar (34) with semantically stylized features (60), and the semantically stylized avatar (34) is output.