Abstract
To enhance viewer engagement, video creators embed links of related content within their videos. Integrating links to related videos is a manual procedure that requires creators to review videos to identify appropriate frames to embed related content and determine appropriate content to be embedded. This disclosure describes techniques that leverage a visual language model (VLM) to automate the spatiotemporal placement of links to related (recommended) content within a video. Subframes of the target video and of potential anchor videos are transformed to embeddings in a shared metric space via a VLM. Candidate anchor videos to link to and spatiotemporal locations for insertion into the target video are identified by obtaining a similarity score in a common metric space. Content creators are provided options to select and insert the identified anchor videos at the suggested insertion locations. Automation of identification of related content and insertion points can substantially increase the speed with which such content can be linked to a video and can help improve viewer engagement.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Shin, D, "Automated Identification of Related Content and Placement of Links within a Video", Technical Disclosure Commons, (September 13, 2024)
https://www.tdcommons.org/dpubs_series/7351