ai AI Video has arrived - It looks incredible

sfoster

sfoster

Staff Member
Moderator
Click this link to watch some of the videos
openai.com

Sora: Creating video from text

openai.com

Really incredible stuff from OpenAI
There was also a wired article, but the article was so terribly formatted with ads I'm not posting it

Instead here are the different videos from the wired article with a direct link.





 
Sort by date Sort by votes
CelticRambler said:
"Adorable dalmation" is literally only part of that input text the software managed to interpret correctly. But hey! look! shiny bright colours! so it must be Really Great Technology (We're All Doomed), right? Even then, the "damn COLORFUL" palette that's impressed you is lifted right out of real life, so how about giving credit to the residents of Burano for their hard work?
Click to expand...

You're right - look at ALL the clips on that web page, and you'll see that the AI only got the first bit right. That said, in a few years, AI may get better, but, then again, pessimists have been saying for half a century that computers will take over.

Hmmm ...... can AI do a movie about people talking about people talking about an interstellar war? 😁
 
Upvote 0 Downvote
Aspiring Mogul said:
You're right - look at ALL the clips on that web page, and you'll see that the AI only got the first bit right. That said, in a few years, AI may get better, but, then again, pessimists have been saying for half a century that computers will take over.

Hmmm ...... can AI do a movie about people talking about people talking about an interstellar war? 😁
Click to expand...
By the time you finally get around to making a movie it'll do a lot more than talking 😄

Obviously it'll get better at following the instructions, but one other thing to consider - in no time flat this will also be generating 3d models, and then you can import them into UE and execute scenes with exacting precision.
 
Upvote 0 Downvote
sfoster said:
You think this old stuff is exactly like the new stuff? like they haven't learned a damn thing? lol
Click to expand...
I didn't say that the new stuff is exactly like the old stuff; on the contrary, I said (or at least implied) that its developers churn out a load of shiny new "content" that looks impressive until you put it to the real-world test; and then you find that, in fact, the developers opted out of dealing with some of the most basic challenges.

I said the five-pawed cat was symbolic, and you've made my point for me:
sfoster said:
Nate is one dude working alone. OpenAI is a multi-billion dollar company at the forefront.
Click to expand...

It was in August 2022 that I criticised Nate's cat animation; in the year and a half since, it's got better. Now we have a multi-billion dollar company publishing a sample that demonstrates they're no better than this one dude - their software fails at one of the most fundamental animation challenges, a challenge that old-tech animators solved a long, long time ago.

And guess who agrees with me? Sora themselves! From their site, they start with the bold claim:
The model understands not only what the user has asked for in the prompt, but also how those things exist in the physical world.
Great! That's a real game-changer. Only then we come up against the caveats:
The current model ... may struggle with accurately simulating the physics of a complex scene, and may not understand specific instances of cause and effect. For example, a person might take a bite out of a cookie, but afterward, the cookie may not have a bite mark. The model may also confuse spatial details of a prompt, for example, mixing up left and right, and may struggle with precise descriptions of events that take place over time ...

Ah, right. So it can understand what the user asks for, except when it doesn't; and it understands how those things exist in the physical world, except when the physical world is complex and includes weird stuff like cause and effect, or left and right, or the space-time continuum.

So it's nothing more than another blunt instrument in the film-maker's toolbox, but hardly a substitute for genuinely creative story-telling.
 
Last edited:
Upvote 0 Downvote
CelticRambler said:
And guess who agrees with me? Sora themselves! From their site, they start with the bold claim:
The model understands not only what the user has asked for in the prompt, but also how those things exist in the physical world.
Great! That's a real game-changer. Only then we come up against the caveats:
The current model ... may struggle with accurately simulating the physics of a complex scene, and may not understand specific instances of cause and effect. For example, a person might take a bite out of a cookie, but afterward, the cookie may not have a bite mark. The model may also confuse spatial details of a prompt, for example, mixing up left and right, and may struggle with precise descriptions of events that take place over time ...
Click to expand...

Pointing out those limitations is a very valid criticism.

My question for you is -- How long do you expect it will take before their team overcomes those limitations? 3 years? 10 years? never?
 
Upvote 0 Downvote
sfoster said:
How long do you expect it will take before their team overcomes those limitations?
Click to expand...
Probably longer than they say but sooner than we might expect.

Even then, I would predict that whatever progress is made will be less impactful than forecast. Just like MS WordArt didn't turn everyone into a graphic designer, Photoshop and iPhone cameras haven't turned everyone into master photographers, and a million journalism graduates haven't filled our news media with Pulitzer prize winners, this'll end up as a resource that is so incredibly expensive it'll be out of reach of small independent film makers or it'll be marketed so agressively ("you can be the next Disney") and so abused commercially that its name will become a synonym for cheap disposable footage (kinda like "Photoshop" is a largely negative term in general conversation) and serious film-makers will opt to shoot their own footage, or commission real artists to work on their CGI elements.

Either way, there'll still be animators working with pen and paper, or modelling clay, or puppets, or 1/8th scale models, or humans in motion-capture suits ... Bear in mind that 35mm camera film and 12-inch vinyl records are still being produced and used today.
 
Last edited:
Upvote 0 Downvote
With this technology, there will be SO MANY INDIE FILMS that the market will be floodded 10x, 50x, maybe even 500x worse than it is right now.
Click to expand...

Gosh I hope so. Could it be filmmaking is about to no longer be a pie in the sky pipedream for a lot more people? Awesome. And hopefully competing tools would be developed besides Sora, with fewer restrictions—by competitors who don't look over your shoulder and police what content you may and may not include in your masterpiece.

Will it mean tools to allow me to do 3D modeling without spending years struggling to learn Blender or Maya etc (unlikely)? Awesome.

Will it mean working with AI to shape a musical score for my film and get it just so when I myself couldn't read or write music to save my life? Awesome.

It's easy to imagine ever more refined tools to shape the product to fix mistakes or to shape it to your hearts desire—eliminating the problem of too many paws, or whatever. Edit and reedit, shape and mold, to your heart's desire. Why not? Maybe that's coming. And maybe it will never be that convenient. Maybe they'll make it cost prohibitive. Time will tell.

Giving Indietalk the ability to make a far-out music video beyond his wildest practical dreams sounds awesome.

Maybe you'd never get the kind of real or nuanced performances from AI acting you can get from human actors—much less the experience of working with real people. Then again, I can imagine advanced AI tools allowing you to edit and shape AI performances until you get pretty much what you want...in as excruciating detail as you will. But just as with traditional and current animated films, I don't see why you couldn't have human actors do the acting part for you. Or, you could shoot a hybrid like they already do today, but hopefully without the 200 million dollar budget etc. And, if you want to make live action dramas with no need for special effects, AI created or otherwise, then awesome.

Might we be on the cusp of your being able, with a PC and a software subscription, to make a AAA or AA space opera spectacular all on your own, well, with lots of help from an AI friend? Totally awesome.

I just watched He Dreams of Giants. What help might have such AI tools been to Terry Gilliam to achieve his Don Quixote dream in spite of such a sadly inadequate budget and all the real world troubles that befell him?
 
Upvote 0 Downvote
Open source Stable Diffusion 3 was announced with "greatly improved performance in multisubject prompts" and just like that Celtic's complaint about not recognizing all the prompt details has been conquered.

It does text perfectly now too!
stability.ai

Stable Diffusion 3 — Stability AI

Announcing Stable Diffusion 3 in early preview, our most capable text-to-image model with greatly improved performance in multi-subject prompts, image quality, and spelling abilities.
stability.ai stability.ai

It's a different company but it's all open source so these solutions are available and coming FAST

I'm actually pretty happy and excited about this now, because I think realistically this is the only chance I'll ever have in my life to make a feature film. I'm not gonna get the money or the marketing budget for a meatspace film, but wow I might actually be able to fulfill a dream before I die thanks to this tech.

Biggest thing I'm looking forward to is when the AI characters start talking dramatically.
It'll be fascinating to see if they are good actors or not, cause most of the actors I work with are nowhere near movie star quality.

My belief is that it's going to replace almost all animators in the near future. Tons of them will lose their jobs, I'm not saying it's gonna replace all regular film, Mogul is the one that chimed in and said it looks good enough to replace live action too although it seems he has reveresed his position on that front for now.

The real rubber meeting the road will be how well the AI can act.
I say it'll replace animation cause you just hire voice actors.
 
Last edited:
Upvote 0 Downvote
indietalk said:
A fun exercise would be to take the older films we made and redo them in AI just to see how they turned out.
Click to expand...
It'd be cool to plug any cartoon in and get a live action too.
How hilarious would that be to plug in The Lion King and get a BETTER version than $200,000,000 one
 
Last edited:
Upvote 0 Downvote
An even better example might be the music industry.

Music used to be $10-$15 an album, if you wanted ONE song, you had to buy an entire album...
Now I have literally millions of songs for free, anytime, anywhere, and artists make all their money from touring now instead of album sales.

Unrelated... maybe he is just spooked, maybe he'll change his mind, but it looks like AI has already removed nearly a billion dollars from the atlanta film industry and ai video ISN'T EVEN RELEASED YET. Give it 5 or 7 more years

variety.com

Tyler Perry Halts $800M Studio Expansion, Citing Concerns Over OpenAI’s Sora

Perry's planned expansion would have added 12 sound stages to the Atlanta backlot.
variety.com variety.com

Edit to add - I had a long conversation with nate last night on the phone, one of the subjects we discussed was AI-Driven NPC in video games.
The gaming industry has 8x more money involved, it's a fucking behemoth, and they are MOTIVATED to get good AI acting.

Like this shit might be giving really good performances a lot faster than we think.
Especially after actors sell their likeness and speech patterns, and we can have a 32 year old pierce brosnan as the star of our films

Don't look directly at him!! He's too handsome, you might go blind.

 
Last edited:
Upvote 0 Downvote
Would it be so bad if this completely GETS RID OF CELEBRITIES?

Celebrities are a blight on our society... people listen to some ignorant ex-child star that never even went to high school, never read a text book, knows absolutely nothing, and society treats them like they're a source of credible information. If this helps put that to an end... it's probably a good thing.

Actors can still have fun on stage.
 
Last edited:
Upvote 0 Downvote
directorik said:
This is going to put a lot of crews out of work.
Click to expand...
AI tech is already hitting the software industry, there's been tons of layoff annoucements and almost nobody is hiring.

According to my dad - for businesses overall in the USA - an increase in 1% productivity equates to 1,000,000 jobs being eliminated.
There's only so much work available to be done, you can quantify that into the inversely proportional ratio.

The entire world is going to shift under our feet, not just the film industry.
Translating that into what this AI tech will be capable of (monkey see-monkey do) maybe making one worker 50% more productive? 80% more productive? across the board. it's very scary. I'm sure those exact proportions won't hold up at high numbers, it'll shake up the system so much.
 
Last edited:
Upvote 0 Downvote
Holy shit - I just put two and two together - I realized I will (one day) be able to use AI to turn Christmas Hellfire in to a VR movie.
It can use the generative AI to completele everything off camera, turn it stereoscopic and I'll literally STEP into my own film and walk around.

It's a real experience that's going to happen to me if I don't die in the next decade or two. Unbelieveable.
 
Last edited:
Upvote 0 Downvote
Looking into this just a little bit I found this YouTuber, Wes Roth. I don't know if he knows what he's talking about. Maybe he doesn't, but he seems to. It's some pretty interesting stuff.

Here, he talks about how it looks like the OpenAI people are using Unreal Engine 5 with Sora. The TL;DR is: they might be using UE5 to create huge amounts of iterations for the AI to make use of.


Here, he explains why the above gives us a solid reason why someone would make a world simulation, perhaps like our world, because they/we are now doing just that. TL;DR: to create more data to drive their AI.



This is going to put a lot of crews out of work.
Click to expand...

And if Wes Roth's speculation in the following video is right then there might be a whole lot more change than that about to occur. When they fired Sam Altman recently, before bringing him back, it was reported there was at least a rumor some in the company were alarmed by something, which led to them pushing Altman out. Might that lend some credence to this speculation? Of course, it could have been entirely unrelated.

Anyway, the TL;DR is: they might already have an AGI in the lab.


While it's unfortunate white collar, or cognitive workers, if you prefer, might soon be replaced by AI, the ugly truth is blue collar workers have been being replaced by automation for decades. The death threats and calls for Butlerian Jihad seem just a bit much. Maybe they should learn to code. Oh wait.

I'm sure we'll all believe it when we see it. Until then…
 
Upvote 0 Downvote
You must log in or register to reply here.
Back
Top