Virtual Object Action Video Splicing for Smooth Frame Transitions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for processing virtual object actions are inefficient due to high time costs and user operation difficulties in generating action videos, particularly in generating action video frames using generative adversarial networks (GANs).
Innovation Solution
A method involving receiving an action instruction with an identifier and time-dependent information, determining an action video frame sequence, identifying a preset state image at a target time, generating a connection video frame sequence to connect the preset state image with the action video frame sequence, and splicing these sequences to create an action video, utilizing pre-collected and pre-stored action videos.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If generative adversarial networks (GANs) are used to generate action video frames, then action video generation capability is improved, but processing time and operation complexity increase significantly
Solution Approach 1:
The patent pre-collects and pre-stores action video frames and their corresponding action identifiers in a database before actual use. When generating an action video, the system retrieves pre-prepared video frames based on action identifiers rather than generating them in real-time using GANs, significantly reducing processing time while maintaining video generation capability
Solution Approach 2:
The patent uses pre-collected action video frames as copies or references to generate new action videos. Instead of creating entirely new video frames through complex GAN generation processes, the system selects and combines existing pre-stored action frames that match the required action identifiers, reducing computational complexity and time requirements
2Manufacturing precision
If key point information is provided according to time sequence, then action video generation accuracy is improved, but user operation difficulty increases
Solution Approach 1:
The patent enables the system to automatically match action identifiers with corresponding action video frames without requiring users to manually provide key point information in time sequence. The system self-determines the appropriate video frames based on action identifiers, eliminating the need for complex user input while maintaining generation accuracy
Solution Approach 2:
The patent introduces action identifiers as intermediary elements that bridge the gap between user intentions and video frame selection. Instead of requiring users to provide detailed temporal key point information, users simply input action identifiers, and the system uses these intermediaries to automatically retrieve and combine the appropriate action video frames, significantly simplifying the operation process
3Stability of the object's composition
If action video frames are spliced according to time sequence, then action video continuity is improved, but processing complexity increases
Solution Approach 1:
The patent divides the action video into discrete segments or frames, each associated with specific action identifiers. The system processes and splices these segmented frames based on their action identifier sequences rather than continuous time sequences, simplifying the processing logic while maintaining video continuity through proper segmentation and reassembly
Data Source
AI summary
A method and apparatus for processing an action of a virtual object, and a storage medium are provided. The method specifically includes: receiving an action instruction, the action instruction including: an action identifier and time-dependent information of performing an action associated with the action identifier; determining an action video frame sequence corresponding to the action identifier; determining, from the action video frame sequence, an action state image corresponding to a preset state image of the virtual object at a target time, the target time being determined according to the time-dependent information; generating a connection video frame sequence according to the action state image, the connection video frame sequence connecting the preset state image with the action video frame sequence; and splicing the connection video frame sequence with the action video frame sequence, to obtain an action video. Embodiments of this application can improve action processing efficiency of a virtual object.