Script-Based Scene Graph Editing for Previsualization Alignment

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

The generation of previsualization videos through text analysis often results in images that are not intended by the user, requiring complex editing of 3D asset arrangements, making the process cumbersome.

Innovation Solution

An information processing device that analyzes a script to extract meta-information, generates previs videos, and allows for easy editing using scene graphs and other visualization tools to simplify the positional relationship adjustments of components.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If text analysis is used to automatically generate previsualization videos, then productivity is improved, but manufacturing precision deteriorates because the generated videos do not align with user intent

Engineering Contradiction:

Improveprevisualization video generation efficiencyVSAvoidalignment with user intent

Core Design Contradiction:

ProductivityVSManufacturing precision

Solution Approach 1:

The system extracts positional relationship information from the script and uses it to guide the automatic arrangement of 3D assets, creating a feedback loop where textual descriptions are continuously referenced to ensure the generated previsualization aligns with user intent while maintaining automated efficiency

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

Positional relationship information serves as an intermediary between the script and the 3D asset arrangement, translating textual descriptions into spatial configurations that bridge the gap between automated generation and user intent alignment

Inventive Principle:

Principle #24Intermediary (Mediator)

2Manufacturing precision

If editing is performed on previsualization videos generated using 3D models, then manufacturing precision is improved, but device complexity worsens due to complicated asset arrangement operations

Engineering Contradiction:

Improvevideo alignment with user intentVSAvoidediting operation complexity

Core Design Contradiction:

Manufacturing precisionVSDevice complexity

Solution Approach 1:

The positional relationship information acts as an intermediary that simplifies the editing process by providing a structured representation of spatial arrangements, allowing users to make precise adjustments without directly manipulating complex 3D asset configurations

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The system creates a simplified copy or representation of the spatial arrangement through positional relationship information, which can be edited more easily than the actual 3D asset configurations while still maintaining the ability to update the original arrangements

Inventive Principle:

Principle #26Copying

3Manufacturing precision

If detailed positional relationships are extracted from scripts, then manufacturing precision is improved, but loss of information increases due to the complexity of processing unstructured text

Engineering Contradiction:

Improvepositional relationship accuracyVSAvoidscript interpretation accuracy

Core Design Contradiction:

Manufacturing precisionVSLoss of information

Solution Approach 1:

The system selectively extracts only the essential positional relationship information from the script, separating relevant spatial data from other textual content to maintain precision while avoiding the complexity of processing the entire unstructured text

Inventive Principle:

Principle #2Taking out (Extraction)

Data Source

PatentEP4697275A1Information processing device, information processing method, and storage medium

Publication Date: 2026.02.18 SONY GROUP CORP

Figure 1
Figure 2
Figure 3

AI summary

Making it possible to output information indicating a positional relationship of components acquired from a script. An information processing device including a control unit that acquires information regarding a component from an input script, performs processing of generating, from the information regarding the component, information indicating a positional relationship of one or more components forming a scene, and performs processing of outputting the information indicating the positional relationship.