Abstract
Generating synchronized contextual information for live or prerecorded media can require substantial manual research and production effort, and viewers may lack timely facts related to what is being shown or discussed. A computer-implemented technique uses a language model with media signals and external knowledge sources to generate informational overlays or annotations relevant to a playback timestamp or live event moment. Audio, caption text, keyframes, viewer-interest signals, and retrieved facts may be used to select, cache, segment, update, and present annotation candidates. Selected annotations may be displayed during playback, surfaced on a companion device, or provided as audio commentary. The technique supports scalable contextual media annotation, personalized presentation, and efficient reuse of timestamped annotation data.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Soland, Michael Paul, "Contextual Media Overlays Using Playback Signals and Retrieved Information", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/12048