Text-to-3D Video Generation Using Parallel Multi-View Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing 3D asset generation methods are slow and inefficient, requiring significant time to generate and edit 3D assets, leading to prolonged user wait times for visualization.
Innovation Solution
A system utilizing two machine learning models to generate and visualize 3D assets quickly by generating multi-view images in parallel, allowing for rapid rotation and editing based on text input, with a predetermined camera offset between image sets.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If existing 3D asset generation methods are used, then high quality 3D assets can be generated, but the generation time is very long (30 minutes to 2 hours)
Solution Approach 1:
The patent segments the 3D asset generation process into multiple independent view angle generations (front, back, left, right views). Each view angle is generated separately by the machine learning model, allowing parallel processing and significantly reducing total generation time while maintaining quality through focused optimization of each segment.
Solution Approach 2:
The system performs preliminary actions by generating multiple view angle images simultaneously in parallel before assembling them into the final 3D asset. This parallel preliminary generation of all required views eliminates sequential waiting time and accelerates the overall process while preserving quality through comprehensive view coverage.
2Manufacturing precision
If existing 3D asset generation methods are used, then detailed 3D assets can be created, but user wait time for visualization is prolonged
Solution Approach 1:
The patent divides the visualization task into segmented view angle images (front, back, left, right views) that are generated in parallel. Users can immediately visualize the segmented views without waiting for complete 360-degree rendering, reducing perceived wait time while maintaining detailed quality through focused generation of each view segment.
Solution Approach 2:
The system creates multiple copies of the 3D asset from different view angles simultaneously. Instead of generating one complete slow rendering, it produces multiple view angle copies in parallel that can be immediately displayed to users, reducing wait time while preserving detail through redundant generation approaches.
3Manufacturing precision
If existing 3D asset generation methods are used, then accurate 3D models can be produced, but editing and modification time is increased
Solution Approach 1:
The patent segments the 3D model into independent view angle images that can be individually edited and modified. Users can edit specific view segments without reprocessing the entire model, maintaining accuracy through focused editing while significantly reducing editing time through modular manipulation of segmented views.
Data Source
AI summary
The present disclosure describes techniques for generating three-dimensional videos based on text using machine learning models. Text and inputting data indicative of a set of multi-view images are input into a machine learning model. Content of the set of multi-view images is associated with the input text. The machine learning model comprises a plurality of sub-models corresponding to a plurality of sets of camera parameters. A plurality of sets of multi-view images is generated based on corresponding camera parameters by the plurality of sub-models. The plurality of sub-models are configured to run in parallel to generate the plurality of sets of multi-view images. A three-dimensional (3D) video is generated based on the plurality of sets of multi-view images. Content of the 3D video is associated with the input text.