Google patents AI video tool that inserts missing brands into ads

1. A method for assessing a video component comprising: receiving, by one or more processors, the video component; evaluating, by the one or more processors, using a multimodal video component evaluation tool, the received video component relative to audio and video metrics, the evaluation comparing attributes within the video component to the audio and video metrics to determine whether each of the audio and video metrics are met by the received video component, wherein the multimodal video component evaluation tool comprises a product/brand engine configured to: receive as input, pixel information corresponding to individual video frames of the received video component, and process, using a pixel-level image models, the pixel information by matching objects identified in the individual video frames to a taxonomy of brands and/or products to identify brands and/or products within the video component; generating, by the one or more processors, based on the evaluation, a result indicating that one or more of the audio and video metrics are not met, wherein the result comprises an indication that the product/brand engine did not detect a particular brand and/or product; generating, by a generative artificial intelligence (AI) model, content representing the particular brand and/or product based on the result; integrating, by the generative AI model, the generated content representing the particular brand and/or product into the video component; and outputting, by the one or more processors, the video component including the particular brand and/or product.