US 12,439,108 B2
Video clip learning model
Amirreza Ziai, Menlo Park, CA (US); Shervin Ardeshir Behrostaghi, Campbell, CA (US); and Eugene Lok, Los Angeles, CA (US)
Assigned to Netflix, Inc., Los Gatos, CA (US)
Filed by Netflix, Inc., Los Gatos, CA (US)
Filed on Jan. 9, 2024, as Appl. No. 18/408,325.
Prior Publication US 2025/0227316 A1, Jul. 10, 2025
Int. Cl. H04N 21/25 (2011.01); G06V 20/40 (2022.01); H04N 21/234 (2011.01); H04N 21/8549 (2011.01)
CPC H04N 21/251 (2013.01) [G06V 20/48 (2022.01); H04N 21/23418 (2013.01); H04N 21/8549 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method comprising:
accessing at least one video clip associated with a media item;
comparing the accessed video clip to one or more previously selected video clips that have been used to promote each clips' corresponding media item;
upon determining that the accessed video clip is sufficiently similar to at least one specified video clip of the previously selected video clips, accessing a take fraction value associated with the specified video clip that indicates how well the specified video clip has correlated to user views of the associated media item; and
based at least on the association between the specified video clip and the take fraction value, training a machine learning (ML) model to predict a video clip take fraction value for the accessed video clip indicating how well the accessed video clip will correlate to user views of the media item.