"Generative AI": Supercharging Humans, Transforming Creativity
Originally Aired - Monday, April 17 | 3:00 PM - 3:20 PM PT
This Session Has Not Started Yet
Be sure to come back after the session starts to have access to session resources.
“Generative AI” can enable broadcast/OTT to transform discovery/viewing across broad content libraries, helping to build audience loyalty/long-term value while reducing the operational complexity. “Generative AI” uses machine learning algorithms to create new content based on “intelligence” gathered through training on a large corpus of text (even code!), images, audio, and video files. These large, pre-trained models can solve domain-specific problems, through transfer learning, without significant computational overhead. In media/entertainment, Generative AI tools and technologies such as Dall-E 2, GPT-3, and CLIP can revolutionize creation of multimedia artifacts – with minimal human guidance - and the way we understand images and videos. Rapid, automated scans of video or audio tracks can go more deeply into the nuances of content, enabling content owners, broadcasters, and streaming providers to quickly and cost-effectively build increased portfolios of supporting video and audio collateral, like thumbnails and key art, that can be targeted at a cohort or individual level. These artifacts can be stored in the video CMS for use in refreshed on-screen catalogs or for recommendations that can be targeted toward specific audience cohorts. Additional posters and supporting images can be created based on the machine’s ability to identify which will resonate well with various audience segments or which is likely to go viral. Using natural language interfaces, a creative professional can generate video descriptors/tags that can be actioned on (e.g. find a scene where a female is in front of the Eiffel Tower, or on a moving train).
This presentaion would show early applications of Generative AI technology in broadcasting and streaming, most notably AI-generated images (or relevant thumbnails from the video) optimized for different target audience cohorts based on their past content preferences and interests.