Using AI to create 3D objects from 2D images is certainly doable, with a number of issues. I am using Claude to generate suitable real-time, web browser based animated, and puppeteer-friendly, characters as part of several larger projects.
This is a real time simulation of a classic, antique doll that my roommate in NY had purchased. It had been used as part of a Christmas display at a major department store. I am not sure which one.
This will run on your web browser. Use 'h' to reveal the controls, 's' to sing, and 't' to talk.
https://ct68-doll.onrender.com/
There are lots of things that could be better, but it already is suitable for its intended purpose. It was basically about 1 day work spread over a week.
That is remarkable.

