Change picture using openai?

Hello - is it somehow possible to change an existing image using openAI?
So i have an image with a face picture of a person.
Is it possible to change the picture to an eg. vector styled 3d animation?
And also made some changes like “make him more excited, make him look serious etc”

How can i do this?
Which API i have to use?
Any examples somewhere?

The short answer: no. Not anything of significant quality at least.

The long answer: What you are describing is within the bounds of what Adobe products are currently specializing in.

However, image → “vector styled 3d animation” doesn’t exist. Something can be a 3D polygon or a vector, but not both. On top of that, animating it is an entirely different process.

Doing the best to break down these steps:

image → vector : Possible with certain adobe products
image → polygon : Possible, but very much in it’s infancy (this usually requires a month’s worth of work by a 3d sculptor). I’ve heard China released an AI tool for this very recently, but haven’t tested it myself.

I’d also note if you’re looking for that pseudo-3d look from 2D imagery, you don’t actually need the pieces to specifically be vectors (although it helps for scaling).

Your two options for professional-grade software for animating said imagery is Live2D cubism and Spine. If you’re using it for gaming, Unity has options here, and there is a free tool that’s commonly used in China called DragonBones, but from purely my own experience and other anecdotal takes on it, it is clunky as hell, and not really worth it. Blender I’ve heard might also have some tools for this, but doesn’t have all the features one would want in comparison to the top two above.

Either way, animating is a process. A long process. The shortest one is using Spine, but it’s also the most expensive.

Trust me, if this was easy and an AI could do it well I’d already have a plethora of video games published by now. Many people would. But this kind of stuff outside of oddly morphed facial animations has a long ways to go before the pipeline gets fully automated.