Consume Product

On the money with this. Although, in some cases, if you supplement with your own labor, you can get much closer to "fine tuning" AI output.

I currently use it to assist my workflow doing digital art. I'll do a rough sketch of a character's composition, then line art, color it, all as if I'm just making the image myself. I try to get as close a possible to what I envision the picture in my head to be. The result ends up mediocre, since I am not a skilled artist, but it's usually coherent enough to convey a clear idea of what I was trying to make.

This is where AI comes in. I feed this image to AI and ask it to give me a professional, finished version of the picture I drew.

Every single time, within 10 image generations it gives me something close to perfect in terms of what I wanted the image to look like. I then take the AI's output and trace over it on fresh layers to clean up my line art, manually edit some details, and then recolor and finish it as my final product (leftist twitter artists on suicide watch.)

It's an extremely useful tool, this sort of workflow is already rapidly becoming the norm.

CalArts BS like "Steven Universe" is easy for AI. One artist can draw the Character/ Model Sheet and then just animate their environments after that. They can train AI on "emotes" of character movements and facial expressions, assign a prompt to each, and then use that as stage direction.

There should be doubt or mistake that Hollywood is working with AI models to replace anyone and everyone they can. The future of filmmaking will be a screenwriter, a director and a producer in a room feeding reference material and prompts into the LLM.

> But the real trick with this stuff is consistency, and camera work.

Correct. What Lionsgate is doing is combining Pre-Visual software where characters and camera movements can be controlled. Basically, the characters can be simple stick figures and a director can plot the movement of the camera and even focus rack for a scene, then allow the AI to "paint in" the characters, environment, etc.

Designers can also create 3D renders of characters, creatures or objects (cars, spaceships, etc.) and place them as "persistent objects" so the AI isn't burdened with "inventing" or rending them with every frame. The only thing that isn't quite there is AI voice, but we've already mastered ADR recording for animated films, so that's no issue. This is going to cut down principle photography down to a week. We're going to be able to make Academy Award-winning films in a week!