How do you think CGI for photorealistic found footage would be optimized? What new methods might be used?