Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Sora: Creating video from text

openai.com

Post reply on HN
Print view View original

311–320 of 1001 posts

Re: Sora: Creating video from text

#311

Post by Krasnol »

chasing wrote: post #24

Yeah, you just can't let all media, all the cost and hard work of millions of photographers, animators, filmmakers, etc be completely consumed and devalued by one company just because it's a very cool technical trick. The more powerful these services become the more obvious that will be. What OpenAI does is amazing, but they obviously cannot be allowed to capture the value of every piece of media ever created — it'll…

Never ever will there be everyone at the table. This is not how the Internet works. It is not how the world and humanity work. If OpenAI doesn't do it, the next big player will. China will. Maybe it'll soon not even need China because it'll be so easy to deploy.

There is no stop now. It's too late for that. Time to think about the full development and how we'll handle that. How we as people will be able to exist next to it. What our purpose in the world is supposed to be. What the purpose of "value" is. What the purpose of "economy" or "the market" is.

Exiting times.

Re: Sora: Creating video from text

#312

Post by pimlottc »

zemo wrote: post #5

> Prompt: Historical footage of California during the gold rush. this is the opposite of history

Yeah, my heart sank when I saw that.

Social media is really good at separating content from context, things like this will distort people's understanding of history.

Re: Sora: Creating video from text

#313

Post by downWidOutaFite »

swalsh wrote: post #153
thepasswordis wrote: post #100

OpenAI demonstrating the size of their moat. How many multi-million-dollar funded startups did this just absolutely obsolete? This is so, so, so much better than every other generative video AI we've seen. Most of those were basically a still image with a very slowly moving background. This is not that. Sam is probably going to get his $7T if he keeps this up, and when he does everybody else will be locked out foreve…

"How many multi-million-dollar funded startups did this just absolutely obsolete?" The play with AI isn't to build the tools to help businesses make money, the play is to directly build the businesses that makes the money. In practice this means, don't focus your business model on building the AI to make text to video happen. Your business model should be an AI studio, if the tech you need doesn't exist, build it....…

But then you're stuck playing in the model owner's playground and if you're too successful they can yank the rug from under you and steal your business any time they want.

Re: Sora: Creating video from text

#314

Post by gwern »

ben_w wrote: post #294
thepasswordis wrote: post #100

OpenAI demonstrating the size of their moat. How many multi-million-dollar funded startups did this just absolutely obsolete? This is so, so, so much better than every other generative video AI we've seen. Most of those were basically a still image with a very slowly moving background. This is not that. Sam is probably going to get his $7T if he keeps this up, and when he does everybody else will be locked out foreve…

> Sam is probably going to get his $7T if he keeps this up, and when he does everybody else will be locked out forever. I would be extremely surprised if he could get past the market cap of all current corporations as an investment. That doesn't mean "no, never"[0], but I would be extremely surprised. $7T in one go would be 6.7% of global GDP, and is approximately the combined GDP of Japan and Canada. > These videos…

Yes, the 'special effects' effect will kick in. Within a year or so, you'll spot this easily, quite aside from the more obvious issues. (That Landrover captioned 'DANDOVER' - is this still using BPEs?!)

Aside from visual plausibility, there's also the issue of physics: one of the things you would like to use video models for is understanding real-world physics and cause-and-effect for planning or learning _in silico_. Something may look good but get key physics wrong and be useless for, say, robotics.

Re: Sora: Creating video from text

#315

Post by bilsbie »

I wish this was connected to chatgpt4 such that it could directly generate videos as part of its response.

The bottleneck of creating a separate prompt is very limiting.

Imagine asking for a recipe or car repair and it makes a video of the exact steps. Or if you could upload a video and ask it to make a new ending.

That’s what I imagine multi modal models would be.

Re: Sora: Creating video from text

#317

Post by jenny91 »

mitthrowaway2 wrote: post #292
jenny91 wrote: post #253

Absolutely insane. It's very odd where the glitches happen. Did anyone else notice in the "stylish woman ... Tokyo" clip how her legs skip-hop and then cross at 0:30 in a physically impossible way. Everything else about the clip seems so realistic, yet this is where it trips up ?

And the cat that wakes up the woman in bed, has three front paws! And that woman seems to be wearing the blanket as though they were pyjamas. Still, it's usually very hard to notice the inconsistencies -- just like the subtle inconsistencies we might see in our dreams.

Yes, there's some really weird hand-blanket morphing going on in that cat shot. Similarly in the guy reading a book on a cloud, the pages flip in a physically impossible way at one point.

I just think it's perplexing how they got things so right, yet so wrong. How did they implement this?!

Re: Sora: Creating video from text

#318

Post by fasteddie31003 »

nopinsight wrote: post #257

Many might miss the key paragraph at the end: "Sora serves as a foundation for models that can understand and simulate the real world, a capability we believe will be an important milestone for achieving AGI." This also helps explain why the model is so good since it is trained to simulate the real world, as opposed to imitate the pixels. More importantly, its capabilities suggest AGI and general robotics could be cl…

Movie making is going to become fine-tuning these foundational video models. For example, if you want Brad Pitt in your movie you'll need to use his data to fine-tune his character.
Post reply on HN