The quality is really superior to what was shown with Lumiere. Even if this is cherry picking it seems miles above the competiton
Yeah we won’t be needing proper raytracing with this kind of tech it’s mind blowing
The second one is easy as you don’t need coherence between reflected and non-reflected stuff: Only the reflection is visible. The second one has lots of inconsistencies: I works kinda well if the reflected thing and reflection are close together in the image, it does tend to copy over uniformly-coloured tall lights, but OTOH it also invents completely new things.
Do people notice? Well, it depends. People do notice screen-space reflections being off in traditional rendering pipelines, not always, but it happens and those AI reflections are the same kind of “mostly there in most situations but let’s cheap out to make it computationally feasible” type of deal: Ultimately processing information, tracking influence of one piece of data throughout the whole scene, comes with a minimum amount of required computational complexity and neither AI nor SSR do it.
This is the best summary I could come up with:
Sora is capable of creating “complex scenes with multiple characters, specific types of motion, and accurate details of the subject and background,” according to OpenAI’s introductory blog post.
The company also notes that the model can understand how objects “exist in the physical world,” as well as “accurately interpret props and generate compelling characters that express vibrant emotions.”
Many have some telltale signs of AI — like a suspiciously moving floor in a video of a museum — and OpenAI says the model “may struggle with accurately simulating the physics of a complex scene,” but the results are overall pretty impressive.
A couple of years ago, it was text-to-image generators like Midjourney that were at the forefront of models’ ability to turn words into images.
But recently, video has begun to improve at a remarkable pace: companies like Runway and Pika have shown impressive text-to-video models of their own, and Google’s Lumiere figures to be one of OpenAI’s primary competitors in this space, too.
It notes that the existing model might not accurately simulate the physics of a complex scene and may not properly interpret certain instances of cause and effect.
The original article contains 395 words, the summary contains 190 words. Saved 52%. I’m a bot and I’m open source!
removed by mod
Ah yes, this definitely won’t have any negative ramifications.
/s
I’m pretty sure that’s a model tho.
I know people have been scared by new technology since technology, but I’ve never before fallen into that camp until now. I have to admit, this really does frighten me.
To make that statement a little more accurate, I’m afraid of the humans that will abuse this technology and societies ability to adapt to it. There’s some amazingly cool things that can come about from this, like all the small indie creators that lack the connections and project management skills to make their ambitions come to life will be able to achieve their vision, and that’s really cool and I’m excited for that, but my excitement is smashed from knowing all the bad that will come with this.
Boo!
What’s wild to me is how Yann LeCun doesn’t seem to see this as an issue at all. Many other leading researchers (Yoshua Bengio, Geoffrey Hinton, Frank Hutter, etc.) signed that letter on the threats of AI and LeCun just posts on Twitter and talks about how we’ll just “not build” potentially harmful AI. Really makes me lose trust in anything else he says.
There with you. This is really worrying to me. This technology is advancing way faster than were adjusting to it. I haven’t even gotten over how amazing GPT2.5 is but most people already seem to be taking it for granted. We didn’t have anything even close to this just few years prior
deleted by creator
This can only be bad for artists and if you are happy about it you are a fascist
The cat video is funny, the cat has 5 legs :D
Seeing the 5 legged cat was the moment I started to believe this stuff really was AI generated.
The folks with access to this must be looking at some absolutely fantastic porn right now!
‘obama giving birth’, ‘adam sandler with big feet’, ‘five nights at freddy’s but everyone’s horny’
possibilities are endless
Honestly, let’s make it mainstream. Get it to a point where it’s more profitable to mass produce Ai porn than exploit young women from god knows where.
I don’t think they would make a model like this uncensored.
Oh its going to be fantastic all right.
Fantastical chimera monster porn, at least for the beginning.
Looks good but still has the ai hallmarks, rotating legs, f’ed up gait… impressive though and it’s going be wild to see what results from this latest pox on the tubes.
deleted by creator
Looking forward to the day I can just copy paste the Silmarillion into a program and have it spit out a 20 hour long movie.
I was thinking exactly this but with the Bible. Not because I like the Bible but because I’d love to see how AI interprets one of the most important books in human history.
But yeha, the Silmarillion is basically a Bible from another universe.
Which is why christians are scared of them. It will open people’s eyes to how anyone can write a fairytale. And so much better ones, too.
If this goes well, future video compression might take a massive leap. Imagine downloading 2 hours movies with just 20kb file size because it just a bunch of prompts under the hood.
Sounds like you already saw Madame Web
This would be the most GPU intensive compression algorithm of all time :)
And the largest ever decoder since it’ll need the whole model to work. I’m not particularly knowledgeable on AI but I’ll assume this will occupy hundreds of gigabytes, correct me if I’m wrong there. In comparison, libdav1d, an av1 decoder, weighs less than 2 MB.
If you randomize the seed it’ll be a different render of the movie every time.
" but you haven’t seen the ultimate limited edition fan version action cut of the directors cut"
I’m really impressed by the demo, but yes, let’s see how well it works when it’s made public.
People who don’t think AI will take a lot of jobs may have to rethink…
