Techniques for analyzing audiovisual content, such as streaming content are disclosed. In one embodiment, a method is disclosed comprising obtaining a frame of audiovisual content, using a video decoder to decode compressed model weights of at least one trained model, using the at least one trained model with the decoded weights to analyze the frame and extract content based on the analysis, using the extracted content to make a determination that the audiovisual content comprises a category of content, and causing actionable information to be transmitted to a client device of a user in response to the determination that the audiovisual content comprises the category of content.