Could long form AI video content be made by transforming acting in 3d engine to photo quality realistic looking content frame by frame?

There are tons of games, 3d environments with excellent graphics. Could this sort of content be upgraded perhaps frame by frame to realistic video?

If yes. Then perhaps a scripting system could be made where a movie is acted inside a game engine according to AI generated script according to a prompt. Then a video would be generated out of that, which then again would be upgraded frame by frame to look real.