Abstract

Online video content platforms host short-form videos in vertical format. Many content platforms also host long-form horizontal videos. For content creators of long-form videos to be able to provide videos in short-form vertical format, they need to dedicate resources to identify appropriate portions of their content and generate cropped short-form videos. This disclosure describes techniques to automatically identify the most suitable segments of long-form content for conversion to short-form content and to generate short-form vertical videos for such segments. The suitable segments are identified using audience engagement metrics provided by the platform that hosts the video content. Object detection and tracking techniques are used to identify the largest object in a segment and to identify a center of the largest object as the focus point. The focus point is adjusted as the object shifts position within video frames of the segment. The horizontal long-form video is cropped using a boundary determined based on the focus point to obtain a short-form video that is free of glitches.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.

Share

COinS