Loading patent…
AssigneeOpenExchange, Inc.
InventorsAlexander J. Kieft, Sean M. Mullins, Joe C. McPherson
Estimating the term…
Computational media editing systems and methods leveraging multi-modal machine learning models are provided. In an example, media files and a natural language editing prompt are received and processed using a multi-modal machine learning model configured to analyze visual and audio content. The system generates structured semantic metadata including scene information, transcription data, object tracking data, and timecodes. A timestamp reconciliation process aligns model-identified spoken content with timecoded speech-to-text data to improve temporal accuracy. Based on the analyzed metadata and the natural language prompt, a structured, non-executable project data object or executable instructions may be generated representing a media editing timeline including segment selections, transitions, effects, and asset references. In an embodiment, the structured project data object may be interpreted by a rendering engine that deterministically generates edited media output without executing model-generated source code. Supplemental media assets may be generated and integrated into the structured project representation prior to rendering.
A computational media editing system using multi-modal language models, the computational media editing system comprising:
(canceled)
A computer-implemented method for automated media editing using multi-modal language models, the method comprising:
Dependencies are inferred from the claim text.