Virtual Animation In-Betweening With Transformer Pose Generation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing animation in-betweening processes are labor-intensive and time-consuming, and automated motion synthesis methods struggle to accurately generate intermediate poses between keyframes, especially for real-time applications requiring exact poses.

Innovation Solution

A transformer-based model is used to iteratively generate intermediate poses between keyframes, employing a bi-directional autoregressive approach with attention masks to predict frames from both the start and end keyframes, enabling efficient and accurate pose generation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If manual in-betweening is used to generate intermediate frames, then animation quality and control over pose-level details are improved, but labor intensity and time consumption increase

Engineering Contradiction:

Improveanimation qualityVSAvoidproduction efficiency

Core Design Contradiction:

Manufacturing precisionVSProductivity

Solution Approach 1:

The system enables automatic self-generation of intermediate frames through a machine learning model that takes keyframes as input and autonomously produces the intermediate animation frames, eliminating the need for manual in-betweening while maintaining animation quality

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The manual mechanical process of frame-by-frame drawing is replaced with an automated machine learning-based system that uses neural networks to generate intermediate frames from keyframes, substituting human labor with computational intelligence

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Productivity

If automated motion synthesis is used to generate intermediate poses, then productivity is improved, but accuracy in converging on exact poses deteriorates

Engineering Contradiction:

Improveproduction efficiencyVSAvoidpose accuracy

Core Design Contradiction:

ProductivityVSManufacturing precision

Solution Approach 1:

The system incorporates feedback mechanisms where the generated intermediate frames are evaluated and refined through iterative processes, allowing the model to adjust and improve pose accuracy based on the relationship between keyframes and generated content

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The machine learning model performs preliminary learning from training data containing accurate pose sequences, enabling it to predict and generate accurate intermediate poses before actual animation production, thus achieving both speed and precision

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS12582906B2System for generating animation within a virtual environment

Publication Date: 2026.03.24 ELECTRONIC ARTS INC

AI summary

The present disclosure discloses the use of machine learning to address the process of motion synthesis and generation of intermediate poses for virtual entities. A transformer-based model can be used to generate intermediate poses for an animation based on a set of key frames.