I can partially answer you. My local film crew friends have a 36 minute "maxi-short" film that is "done" and hitting the festival circuit, finally. We seriously started the thing during Covid lockdown per pre-production / set building. Anyway, we did a lot of green screen for what was to be the last 20 minutes of the film and ultimately show our big, bad, Lovecraftian beast. Our CGI friends over in the Ukraine (yeah, you can already predict where this is going) were on retainer to animate / bring to life said beast. Well, sure enough when the film is done per all the other that's required...our CGI folks were hard to even get a hold of, and first and foremost, they are alive...that's most important...we were way more worried about that. Still, uprooted completely and without any of their equipment or means to replace it at all, we had to make the hard decision to cut most of the beast shots and rely on that you do see / know via bits and pieces that something massive and horrible is right there (coming through the veil / ripping through into our world). It's rather psychological more so, now, but it does still work quite well.
Long story short --- we looked at what AI options allow us at our self-financed level of making this film. There do seem to be some ways that we could do it ourselves, but it's going to take basically still becoming CGI / visual FX artists overall because there's a lot of understanding for coding, CAD, etc. that will be needed. My practical FX capabilities could possibly help by doing everything in miniature and then layering it into the finished film footage, but that's not a full on guarantee, plus it would still cost quite more and take more time and basically keep pushing back the film from getting seen.....especially when the primary point of this "short" film is to hopefully find funding for completing what we have as a longer, full-length film.
On our level / the micro-indie-level, I do think that AI can be of great help, but there's going to be the realization of that old phrase, "one way or another, you pay for your education." Much time and finances and effort will cost you, and more than anything, it's going to require folks to know / decide on what you want your CGI beast/action/concept/visuals absolutely are and exactly what they are doing...in full. Because it's still so hard to get exactly what you want as the finished product, you need to figure all of it out, first, lock in on the visuals as finished and what's 100% going to be there...and then finish out your shots with live action around it all. For me, someone who always writes / composes projects (especially screenplays / prose) backwards, this won't be as hard to do. I think it might give most folks real fits, though.
I do see that of late, you can generate just one AI image and then allow (sometimes the same) AI to animate it for up to 8 seconds in supposedly exactly as you tell it to - even camera panning and such. I'm not up to speed on the quality, but it's interesting because it could make all of the previous backwards talk I mentioned become not the case at all. The technology is certainly making strides. None of it is free, though. You can make a few free mid-res pics here and there, but solid, hi-res and film quality pics and definitely any animated video use are going to cost you a decent monthly subscription fee. If you have the budget, cool. I'm not sure everyone will. It looks to be akin to basically taking on a small car note for the full quality and no limitations for use.
Dang. I guess that actually became long story longer, haha. Oops.