This is the worst it will ever be again.

All the little goofs people caught are the seven-fingered third hands that y'don't really see anymore. They'll mostly just disappear. Some could be painted out with downright sloppy work, given better tools. You want a guy right there? Blob him in. These are denoising algorithms. Ideally we'll stop seeing much of anything based on prompts alone. You can feed in any image or video and it'll remove all the marble that does not look like a statue. (Someone already did 'Bad Apple but every frame is a medieval landscape.')

We're in the Amiga juggler demo phase of this technology. "Look, the computer can render whole animated scenes in 3D!" Cool, what do you plan to do with it? "I don't understand the question." It's not gonna be long before someone pulls a Toy Story and uses this limited tech to, y'know, tell a story.

Take the SG-1 comparison. Somebody out there has a fan script, bad cosplay, and a green bedsheet, and now that's enough to produce a mildly gloopy unauthorized episode of a twenty-year-old big-budget TV show, entirely by themselves. Just get on-camera and do each role as sincerely as you can manage, edit that into some Who Killed Captain Alex level jank, and tell the machine to unfuck it.

Cartoons will be buck-wild, thanks to interactive feedback. With enough oomph you could go from motion comics to finished "CGI" in real-time. Human artists sketch what happens, human artists catch where the machine goofs, and a supercomputer does the job of Taiwanese outsourcing on a ten-second delay. A year later it might run locally on your phone.