As far as I know, the best options for creating ‘scenery’ as a background for film production are being done with ‘text to video’ prompts. Does anyone know if any cutting-edge technologies are being developed? Using text prompts is a challenging task of crafting just the right wording. As I understand it, the technology is getting closer to AI being able to decipher an actual picture. This would be a much easier solution to creating backgrounds that could make the large LED screens used in ‘Batman’ and the like become obsolete.
“In February 2024, filmmaker and producer Tyler Perry announced that he was indefinitely halting a planned $800 million expansion of his studio in Atlanta, Georgia. Perry cited the rapid advancements in AI-generated imagery and video—specifically OpenAI's text-to-video model, Sora—as the primary reason for the decision”.
If the next generation of Sora, or similar AI, can produce a realistic background from still images or existing video, the entire film production will be revolutionized. Films can be produced on sound stages with live actors at a fraction of the cost. Furthermore, if the technology can be devised to scan and track the actors on the sound stage, the AI could potentially create a virtual scene. For instance, if the actors run toward the back-stage wall, the background could potentially adjust to the depth perception. If the sound stage is large enough, this kind of AI-generated background could replace CGI.
Any thoughts or updates about this?
1 person likes this
Michael Joseph DeRosa There's a lot to talk about in your post, but I will be brief. 1. Tyler Perry's announcement was probably a way to save face when reneging on his promise to build an $800m studios in Atlanta, and his celebrity stock (so to speak) went down everywhere because of it. His comments were taken as disrespectful to artist and indicative of a greedy character who never really cared about the local community. 2. AI can be used for environments and backgrounds, but no serious studio will do it off the shelf. The issues are serious copyright problems, and unoriginality - if Sora or Kling or Runway (etc). are generating an image or video for you - they have done ti already for someone else or will shortly, and it's based on what it has been trained on. So what we have to do is create our own reference materials - often image and video, with real locations and real actors, and have the AI system manipulate those materials. This has to fit into a regular VFX pipeline. So you can see, putting in some words and pressing the render button doesn't make anyone an artist or a filmmaker...