(i) a controller configured to perform digital image processing by using an AI model trained to perform data fetching prediction upon at least one input image generated by a static 2D computer generated imagery (CGI) and to create extrapolated output dynamic images that contain more visual data than the at least one input image generated by the static 2D CGI, the data fetching prediction by the AI model extracting a tile from a cache memory before detection of any indication of user viewing interest; and

(ii) at least one display means configured to present the extrapolated output dynamic images to at least one user, the extrapolated output dynamic images containing more visual data than the at least one input image by including the tile extracted by the AI model from the cache memory before detection of any indication of user viewing interest,

wherein the at least one input image is generated offline by the static 2D CGI prior to the data fetching prediction and the extrapolated output dynamic images are free viewpoint 3D images and wherein the data fetching prediction has a reduced-latency and results in production of the extrapolated output dynamic images that comprise novel images as well as novel multi-directional and image scenery parameters in comparison with the at least one input image generated offline by the static 2D CGI prior to the data fetching prediction.