Sora by OpenAI is the Text-to-Video AI Model Beyond Our Wildest Imagination

[btarunr]

Sam Altman of OpenAI just unveiled Sora, the all new speech-to-video AI model that exactly the way science fiction would want such a thing to work—imagine fluid, photorealistic, true-color video clips based entirely on text prompts. Sora is generative AI on an exponentially higher scale than Dall-E, and presumably requires an enormously higher amount of compute power. But to those that can afford to rent out a large hardware instance, this means the power to create a video of just about anything. Everything democratizes with time, and in a few years, Sora could become the greatest tool for independent content creators, as they could draw up entire worlds using just prompts and green screens. Sora strapped to a mixed reality headset such as the Apple Vision Pro, is basically a Holodeck.



View at TechPowerUp Main Site | Source

 

[Solaris17]

and in a few years, Sora could become the greatest tool for independent content creators, as they could draw up entire worlds using just prompts and green screens.

nah no need to wait

this is every south western town iv ever driven through

 

[thesmokingman]

The best pchopper this side of the Mississippi...

 

[Crackong]

Just curious how much computing power it needs to generate 1 frame.

 

[mechtech]

like this one

uh-oh

 

[qwerty_lesh]

Some of it in action:

http://imgur.com/EEyb7EG

http://imgur.com/wpPPQYY

from:

http://imgur.com/a/Om4bFpu

 

[thesmokingman]

lol just read Andre Karpathy left OpenAI. That's gotta hurt.

 

[R-T-B]

I'm reminded of this amazing AI video:

 

[R0H1T]

nah no need to wait

View attachment 334837

this is every south western town iv ever driven through

What's that creature on the far left, first on the road? A misplaced alien? Only see 2 legs there

Some of it in action:

http://imgur.com/EEyb7EG

http://imgur.com/wpPPQYY

from:

http://imgur.com/a/Om4bFpu

The first one reminds me of Ghost in the shell, an absolute masterpiece for its time!

 

[bug]

Imagine being a historian 50 years from now and trying to figure out genuine sources...

 

[Easy Rhino]

like this one
View attachment 334841

uh-oh

This is exactly what I first thought of... LOL

 

[Random_User]

Just curious how much computing power it needs to generate 1 frame.

How many kilowatts being wasted wasted, to render this useless garbage? This has no use, outside to substitude reality. So many power that could be used into schools and hospitals, even keeping the room warm in cold winter day is more useful, than burning watts for this shit. The worst thing, that first crypto, and now AI, made the electricity much more pricey, now it's goint to raise prices globally. Especially due to such greedy ars*es as JHH, who wants this desease to spread worldwide. Sorry, you can cr*p on me whatever you want, but this is my take.

like this one
View attachment 334841

uh-oh

“Were you listening to me, Neo? Or were you looking at the woman in the red dress?” "Morheus"

Sadly, people didn't made any conclusions, and deliberately became the ones that are too late for saving. They won't even get to the point of blue/red pill.

Imagine being a historian 50 years from now and trying to figure out genuine sources...

It's already flood of garbage info and ads. google search is an utter cesspool, and doesn't give any sufficient result, for like couple years already. Have to use other sites and resources manually, to get some decent info.

And your point is spot on. I couple years, the all info that resides at web now, is already can be treated as compromised. That includes info, that predated the "official" AI/GPT stuff, since it removes that thin line of credibility, of when it's all started.

 

[Wirko]

I want exactly the opposite of that: youtuber-to-text technology that can digest a 20-minute video and produce a well-organised 500-word article with a few pics and headings. That should in most cases be enough to retain all the useful knowledge.

 

[R-T-B]

Imagine being a historian 50 years from now and trying to figure out genuine sources...

Just look at the hands. 99% of the time AI can't figure them out.

 

[Prima.Vera]

But still the images are looking fake like they are in a computer game...

 

[Fourstaff]

We are rapidly approaching a point where its basically impossible to tell real from AI, I wonder what will happen by the end of this decade. Desperate attempt to preserve what real looks like, against the unlimited amount of generated content?

 

[Wirko]

Just look at the hands. 99% of the time AI can't figure them out.

But it's often clever enough to blur the face if it's not sure how many eyes a human has.

 

[AusWolf]

First, fake pictures, fake Facebook profiles, now fake videos. At this rate, the whole internet will be fake in 10 years.

 

[cvaldes]

First, fake pictures, fake Facebook profiles, now fake videos. At this rate, the whole internet will be fake in 10 years.

Ten years?!? Try three.

Ahahahahahahahahahahahahaha!!!!!!

The Internet is already full of fake or AI-authored content. Just search for anything online. This has been going on for several years, not just the last few months. 'Bots have infested pretty much every web forum and certainly every single social media network.

CEOs from formerly reputable media sites have been fired for allowing this stuff to happen on their shift.

But this shouldn't be surprising to any Web veterans. The Internet plateaued around 2007-2008, pretty much stayed level for about five years and has been declining since.

Even the once mighty Google search engine doesn't seem to able to discern authentic content and AI processed puke.

 

[bug]

Just look at the hands. 99% of the time AI can't figure them out.

Which is why I've said "50 years from now"

 

[ThrashZone]

Just look at the hands. 99% of the time AI can't figure them out.

Hi,
Feet walking also pretty obvious unless she is wearing cool new roller blade high heels hehe

 

[mashie]

Just look at the hands. 99% of the time AI can't figure them out.

Same with pretty much any text.

 

[remixedcat]

ai generated shit should have been forced to have a marker of some sort cuz it's getting better at hands even motionleap and picsart phone app ai has gotten better at hands

 

[KLMR]

Shouldn't the "news piece" come with a link to an actual video?