Let’s break this down into some separate ideas.
Let’s first consider what we mean by AI generated for this discussion. My definition :
A system that creates from user input an interesting and artistically inspired production.
How that is done is not constricted by that definition. It could be big data, or rule based, or a genetic evolutionary system, or some other approach, or hybrids of the preceding. Please expand on this methods list, if you’d like.
It would seem useful to be able to involve iteratively with this generation system. To have it make a pass, and then be able to go into that pass and give further instructions, and edit, to generate a new pass, that is available for further instruction and editing.
It seems in method it could be based more on an animated :
- 2-D Image generating approach
- 3-D model animation approach, including view animation, and effects.
One way to do a 2-D image generation approach is to generate a series of single frame AI generated images and applying some sort of method to generate the ‘tweens. Things like image morphing, or geometric distortion, or other “tricks” could be applied.
One way to do a 3-D model animation, is to AI generate 3-D models that can be used by already existing 3-D animation software. Then the motion and interaction of these modeled objects, could either be also AI generated, or produced in a more conventional manner, with the final animation done by the already existing 3-D animation software.
It is also possible to combine the 2-D and 3-D approaches.
=-=-=-=-=-=-=-=-=
As an imagined vision of a system like this i recently posted this :
In a few years there will be a Live Action 3-D Animated AI Art Synthesizer, with voice control :
”Ten hula girls dancing in a circle, in the style of the Simpsons, Exploding into a multi-colored amoeba with swirling patterns, That shifts to a Peter Max sunset with sailboats and rocket ships …”