FULL RELEASE of SORA! OpenAI Begins New Age of AI Video | UNLIMITED $200 version tested.

[@OtoGodfrey]

Absolutely wild, the competition is heating up.. don't forget how recently we were just shocked by OpenAI Dall-E image generator. When Dall-E came out, people thought this generation video is 10 years away.

[@thenextension9160]

yep, coherence and hollywood-tier generation is probably only 2-3 years away at best.

[@jenkem4464]

@@thenextension9160 "hollywood-tier" stuff is already pretty much here, at least for small scenes and certain parts. Background and static environment shots, smoke, fire, etc on small scales is frighteningly good if you're patient. Control...well...yeah that's not quite there but man there are some exciting control things popping up. I would say 1-2 years. This is exponential.

[@Slowp0w]

In 10 years you can probably generate your own video adaptations of your favorite novels.

[@konradklimczyk9104]

​@@Slowp0win 10 years we will be in constant stream of dopamine hit from taking part or playing them as a character...

[@MrJetMango]

200 USD thank you but no

[@User-x9x4u]

"his head is on backwards, but other than that..."

[@mydyingdream]

this is "new age of ai video" XD

[@sphanselman6]

Best comment ever!🎉

[@ChrizzeeB]

Not a dealbreaker

[@Vivaildi]

Will never pay for that lmao 200$ for just longer videos basically. And better graphics. But still same glitches as other ai video generators. What a scam.

[@FilmSpook]

@@Vivaildi I hear ya. Mochi, which is open source, has graphics that are just as good as Sora, in many cases, I'd say.

[@TheBroligarch]

In January this was unthinkable

[@Dominik-iv6ek]

Which year?

[@lkrnpk]

This year

[@Dominik-iv6ek]

@@lkrnpk Then I strongly disagree... In January it was nearly obvious that generative video will be a thing.

[@jazionpurnsky1185]

No it wasn't.

[@kristinabliss]

Nice presentation 😊 Thank you!

[@dm204375]

The reason the low frame rate on the reflections of water and puddles is because its interpolating from the many videogame videos Sora was trained on. Many videogames limit the frame rate in reflections to save on gpu resources, that carried over to Sora as thats the source material for some of the training videos. Same goes for that dragon over the city and other videos.

[@MarioCalzadaMusic]

You mean that actually it’s getting it’s images from video games

[@fitybux4664]

Wow that's dumb. I noticed some of the reflections in this video were actually high frame rate reflections. I guess they just need enough data to say "this but not that". (Does Sora have thumbs up / thumbs down?)

[@Gangstor]

@22:58 ...Luigi at McDonald's! 🤣🤣🤣

[@JustinArbabi]

😂😂😂

[@mawns]

All videos are in slow motion for no real reason other than making short videos appear longer. The 5 second video is actually 3 seconds.

[@ThePodmeister]

16:55 my guess is that the low FPS of reflections and shadows is caused by over-training on video game footage. A lot of the time, shadows, reflections, time of day transitions and so on are not rendered at full frame rate.

[@robertfoertsch]

Excellent Analysis, Deployed Worldwide Through My Deep Learning AI Research Library, Thanks Wes ❤

[@jViP70]

200... Damn! Thank you Tim for this.

[@Vivaildi]

you can't even control the camera movmements, that's ridiculous.

[@joMojojojo]

Just imagine what it's gonna be in a year... in 3. By this time we won't be able to believe anything we haven't seen with our own eyes 😂

[@Axiomatic75]

We haven't been able to believe anything we haven't seen with our own eyes for many years now.

[@rolestream]

The internet might be dead by then.

[@Outplayedqt]

@@Axiomatic75 You’ve completely omitted half of America, like my mother, who‘ve fallen prey to misinformation and disinformation.

[@ВладГромов-ь4е]

What if we hit the ceiling?

[@halnineooo136]

Back to the village gossip

[@daniloruffino9093]

Thank you for enabling the audio track

[@dreamphoenix]

Thank you.

[@petratilling2521]

This video should be called “other than that…”

[@AnaLissansky]

I'm stoked! Can't wait to create custom clips for my reels and stories. Hungrily watching all the "so far" tutorials on it :-0 would love for you to do a comparative with Kling AI!

[@gbbenner9382]

Some of these videos make me laugh so much, especially with your narration 😂

[@picksalot1]

I think the colors are outstanding, and add significantly to the videos looking real.

[@nicheva417]

SORA should make itself the site where people upload and host their inevitable short films.

[@SimonWilliams0]

Hey you featured my Korat cat video 😊

[@almor2445]

Not available in the uk! :(

[@kamikaze.33]

3:30 "That guy's had is on backwards, but other than that..." 😂 You say that like it's such a minor flaw! 🤣

[@RobbCochran-l2u]

I did one once "Mike Tyson fighting a T-Rex in a Space Boxing Match on the Moon hosted by ESPN Moon Edition and George Carlin eating a bowl of Oversized Fruit Loops in a Time Machine" ... the outcome was pretty hilarious

[@tomoki-v6o]

Will be very helpful for cinematography students .Great project

[@brjohow]

not really, they wont learn anything about actually making a good film and outsource it to prompting to get aislop.

[@kristinabliss]

Anyone can present their stories with animation now.

[@antman7673]

There definitely is some mastery to it currently. It is still very hard to describe a movie scene for myself.

[@tomoki-v6o]

@@antman7673 i meant for example studying the history of cinema and different techniques just by prompting and seeing what it looks like (interactively),better than theoretical lectures . of course the technology is at its infancy.

[@ВладГромов-ь4е]

There is enough money for students.

[@aifascinated]

Hey can't wait to get my hands on it!! But I'll have to wait a bit because i'm in Europe 🤷 Thanks a lot for all the infos ❤

[@TheReferrer72]

use a VPN

[@aryabrown7904]

"We're not going to replay that."🤣🤣🤣 14:41

[@Tracey66]

I saw Matt Wolfe’s “Monkey on Roller Skates” in there. 😊

[@jasonjackson216]

Not in the UK yet :(

[@tiagotiagot]

I wonder if the reflections going at a lower frame-rate is an indication they trained on game-engines; that is a somewhat common trick to render reflections in "real-time" in games by reusing the results from previous frames instead of rendering fully every frame.

[@mawns]

The reflection is not a different FPS. It is just one model creating the image and its reflections.

[@fitybux4664]

I think the commenter above you had it right. Video games did this trick of "low FPS for reflections", and we are seeing unfortunate artifacts of training on that.

[@mawns]

@fitybux4664 Possible. But first I'd like to see someone count the fps in the main object vs its reflection, because I don't see it.

[@flexibleaspect]

Just out of curiosity, how much did it cost you to make your example videos?

[@fitybux4664]

X / 500 * $200/mo. (Assuming you are factoring the $200/mo as only for Sora and not the o1 Pro model improvements, etc.)

[@SlowlyMakingItHappen]

Super awesome. I can cut my marketing budget down to almost not needing to hire at all now. Last month was already cutting a person with the GPT models. Wow. Mind blow.

[@pablogrenfell7919]

That’s sad

[@SlowlyMakingItHappen]

It's all contract work. And people doing contract work are used to consistent change. These tools are cool and empowering and people are figuring out how to make that trade off. Its not that sad, people have a chance to change their standing in the world and make really good products.

[@jacobbeheler324]

Wow I’m early this time 🎉

[@reinerzer5870]

Everything looks like dream sequences. That actually makes sense.

[@annonymbruger]

Guess most people do something they quickly forget about. It’s insane how many resources that goes into this. Focus should be on more sustainable solutions. This is like throwing hot water out the window.

[@dimamakesslots]

i like the surrealism of it.

[@mikeha]

how do you access sora though? I don't see anything in the chatgpt interface that lets me get to it.

[@RichardHartness]

Regarding the FPS difference in the reflections, if I had to guess, they generated a ton of footage in Unity and when they generated the content for and through Unity, they probably had reflection settings set to a lower FPS so that the ray-trace performance could be higher. If that's the case, this is possibly generated by just one model, however, the model was trained on choppy reflection data.

[@fitybux4664]

They'll need a way to thumbs up / thumbs down specific parts of the video. Like a way to paintbrush areas. Give it a thumbs down on the bad reflections from crappy video game rendered reflections, thumbs up on real world reflection videos.

[@remaincalm2]

I came here for the guinea pig fight. Why didn't you show it? Can we have a rematch please? 😂

[@luckybrandon]

Does it do consistent characters? If not which platform does?

[@aspzx]

No it doesn't. There's an example of a news clip on AI Explained where the image is very convincing but the text chyron is all mangled.

[@mutantdog.]

Wes, you need to give it an image of a research paper and tell it to scribble over various words in bright colours.

[@user-qw1rx1dq6n]

7:33 oh my god he’s heading for the tower

[@fabianoperes2155]

This timeline prompting make all other video AI look old already. This looks so usefull. Why didn't anyone think of this before?

[@fitybux4664]

Think about what it is actually doing to achieve this. It seems almost like magic.

By the trend on how often they act like they released something while releasing it only in US I start to suspect they have some background deals with VPN companies.

[@pantherasad2114]

it looks very impressive!! But I think some of the competition have caught up to Sora🤔

[@parthasarathyvenkatadri]

I think they should have a way to feed the generated videos back into the AI and then ask it add features like the train moving etc

[@Baleur]

8:46 am i the only one that gets absolutely freaked out by how similar these AI videos are to our own dreams? Or am i unknowingly an android? Becuase my dreams often have the exact same illogical "errors" where a street corner turns into an apartment window, or someones face morphs into someone else. I recognize "flaws" in the AI videos as memories of some weird dreams ive had myself.. Now obviously this has nothing to do with suggesting Sora or similar AI's being "conscious" or alive. But its just very very peculiar, how such relatively primitive "dead" algorithms can create outputs that are so extraordinarily similar to our own brains internal dream state. Especially the flaws they manifest... The human brain is completetly different from the popular trope "its just a big computer", its completely different from any computer. But still, our own internal "videos" when dreaming also often feature the exact same sort of twisted flaws. Very very peculiar, and kinda uncomfortable.

[@PoffinScientist]

I'd be amazed a couple of years ago to see early video generation technology to be this good so soon. I don't think, however, that brute force harder training will improve so much from this point, I'm guessing we're almost reaching a plateau and it would take insane amount of computing to improve considerably. For some time I have believed that, first step, the AI must sketch a simple framework of a coherent environment and objects, then, on top of that it should do the hard work of video generation, corresponding to the framework.

[@jenkem4464]

That's pretty much the needed step that is being put in place. Midjourney's retexture and all the other control stuff using depth maps, normal maps, line/edge detection, shape detection, etc are all heading in that direction. Kling ai's motion controls are quite nice and work to give you some rough control over the general movement of specific parts of the image.

[@mydyingdream]

Leaving subscription on this channel, I'm constantly shocked by something that's just an average news. IT'S NEW AGE OF VIDEO, which means we have another not too bad video generator.

[@Outplayedqt]

The difference being, this isn’t like flux.1 or Kling, where you need a tiny bit of effort to get going. OpenAI’s interface has made this incredibly accessible to anyone who has access.

[@benjamincastle3469]

Im shocked that you're shocked that he still says shocked. He's explained its an algorythm thing. For every person he loses, he probably gains another 20 by using 'shocking'. Why not just think "Ok, thats a thing he does" and move on. It doesnt effect your life in any way.

[@mydyingdream]

@@benjamincastle3469 That's exactly what I'm doing. I wasn't on this channel for months. And sometimes I just feel like commenting on things on the internet, that's what comments are for, isn't it? ;)

[@benjamincastle3469]

@@mydyingdream ok. Fair enough. You may not be aware in that case. Its an algorythm related thing. He got lots of dislikes for doing it. Eventually it kinda became a (semi) in joke and now he keeps it up, in part, for that reason. Genuinely, have a good day man.

[@mydyingdream]

@@benjamincastle3469 oh, nice. Clever move then. In terms of money making of course. Thanks for clarifying!

[@nicheva417]

They need the eleven labs audio feature.Just link it to the clip.

[@lj9513]

"He looks VERY intense.... he looks like he's about to lose his mind." hahaha

[@fitybux4664]

It looked like a normal type A professional worker. Maybe that doesn't disregard that people who are in this category do always look like they're about to lose their mind? 😆

[@PRESTONThornburg-u2w]

Web3 Infinity Token's technology is ahead of its time. This project has massive potential!

[@densonsmith2]

How many years do I have to wait for my use case of "effectively a simulation"?

[@E.Hunter.Esquire]

It's only worth it if it makes the kind of videos we all know we want 😂

[@robelsac]

Shhhh, we don’t talk about sex. We’re beyond such grotesque things.

[@fitybux4664]

If it reads your ChatGPT memories, it might be able to infer more context of WTF you mean, and add some clarifying parts to the end of your Sora prompt. For some reason, people make SOOOOOooo many assumptions with this thing already, as you can see in the sample videos.

[@juanmiguelchavezarevalo4667]

It’s crazy how fast this people achieve this AI it’s ridiculous, I can’t imagine how 2030 it’s going to looks like

[@fitybux4664]

No more humans. Easy to envision.

[@PRODIGY5369]

It's so impressive. It just makes me instantly worry about the future. Currently there are so many amazing things you can see on video which are real and recorded, actual genuine history. In very short time we are going to be flooded with videos that will be impossible to tell if they are real or not. Some people might say, so what? Who cares? I do ! When the AI aces this, and is undetectable - nothing that we see can be reliable, trustworthy or accurate.

[@AyselBurhan-pn7ud]

Web3 Infinity Token is gonna pop anytime soon

[@thecoffeejesus]

Google just announced Willow, their new quantum chip. It is as reliable as traditional chips. Error rates are steady and scalable. It’s scalable and reliable.

[@StratumPress]

"Google" lol

[@handsanitizer2457]

Try out the new Gemini in the ai studio easily one shots things ​@StratumPress

[@Douchebagus]

Now if only there were good uses for quantum chips.

[@aspzx]

"Reliable as traditional chips" it can hold its internal state for 100 microseconds. Impressive for a quantum chip but I wouldn't call that as reliable as a normal chip.

[@lupusreginabeta3318]

Sora doesn’t cost 200 dollars to use you can use it for 20 dollars

[@reinsein518]

bro why the cliffhanger why

[@Kram1032]

"but other than that, this looks great"

[@SaintFort]

In a few decades, movies and television shows will be able to be made completely by AI (at least visually) and actors and cinematographers will become obsolete.

[@ReginaCæliLætare]

0:58 CEO killa

[@corysteiner1218]

I've noticed that sci-fi related creatures, like dragons and aliens for example, are consistently very low resolution models and unrealistic. Their physics are way off as well. Not sure if that was intentional by SORA, but that is a dock off their score for me. However, models that are based on reality, such as a dog walking next to its owner on a snow filled sidewalk, works fantastically.

[@Experternas]

why is it called Swedish Sauna in america? We don't have saunas, we never had. no one uses a sauna in sweden. it's a finnish thing. Finland uses saunas and in Sweden it's often reffered to as a Finnish sauna.

[@TheFakeGooberGoblin]

Maybe you don’t have one… they have em in Texas bro. Anyone can own a sauna, unless yall got a ban on it. We SaunaKeepin now?

[@nicoli3143]

Because they were testing if openai would force it to make a sauna full of africans

[@Experternas]

@@TheFakeGooberGoblin I didn't say no one can own a sauna, im talking about the expression "swedish sauna" and saying we don't have saunas in sweden, its not a thing. It is however a thing in Finland where every dude has 5 saunas in every room of the house. The expression should be Finnish sauna

[@TheFakeGooberGoblin]

@@Experternas I think your misconstrued on what associates some thing with a country. If they made Japanese katanas in America based on incorporating American standards and traditions into Japanese design, now they’re American katanas based on how they were made by Americans in America even if every American household doesn’t own a katana. Therefore a Swedish sauna doesn’t imply everyone in Sweden uses saunas, but that it is a company based in Sweden with an appreciation for saunas. If it was called “American Tea” I wouldn’t go “But Americans drink coffee, that’s Britain that drink tea why not call themselves British tea?” I would assume it was a company that loves tea and is based in America. Get what I’m saying? What’s this mindset that to put any product next to a countries’s name, the entire country must first use that product regularly? Wait til you hear about British Petroleum when Americans invented the oil rig.

[@Experternas]

@@TheFakeGooberGoblin no no :) Nevermind. I guess that blew right over your head.

[@Sajuuk]

Not affordable for private use unless you earn a good salary. It's worth noting that most of the AI revolution is passing by hundreds of millions, if not billions of humans too poor to pay subscription fees. Of what use is AGI and more importantly ASI to people unable to access it? The First World will benefit from it first while developing countries' citizens may only benefit from it until decades later if at all.

[@AaronSof]

II cant get it in the uk!!!!!!!

[@lavafree]

physics still so screwed 😂😂🤦🏼‍♂️

[@ytpah9823]

unbelievably underwhelmed with the output. technically blown away by what it took to get this far.

[@nicoli3143]

Still a long way to go... Maybe a year

[@RafaelMilewski]

It feels worse than the ones that has already been released for months....

[@Wizardess]

24:42 -watch the feet on the monkey in the center frame. Its right food becomes its left foot as it swings around from in back for the next step. {^_^}

[@midiwhale]

"His head is on backwards, not too bad" - hardly the video killer it was hyped up as being :-( How will it learn to grow?

[@peterschlange1832]

Grok 3 is gonne be out of this world! Its being trained on a cluster 9x more powerful then the one chatgpt 4 trained on.

[@DanFrederiksen]

Looks half decent but it needs persistent characters and the ability to deliver lines including voice audio. And it needs to be able to generate known people. No hysteria is welcome. It can just be specified that there are no rights for any likeness to be used commercially, obviously. But it cannot have hysterical limitations or other models will be used. If I want it to depict biden, blinken and netanyahu at the hague it better well generate that. Or Schwarzenegger as Dirty Harry

[@youngop]

Is there a reason you didnt use your credits to generate anything for us?

[@LewisDecodesAI]

10:56 Oh dear. Imagine

[@parthasarathyvenkatadri]

Hmm i believe we are going to start having AAA games that are generated on the fly with a new type of hardware that is specifically made for AI workload

[@ZsoltSzentirmai]

It seems it"s still only good for generating dreamlike, fantasy (hallucination-like) scenes.

[@patricknelson]

Still very surreal and nightmarish. Don’t get me wrong, technologically it’s pretty amazing. But it still feels very dream like and has some coherence issues. Like the stampede of bears with individual bears fleeing in and out of existence, merging into one another, with some disappearing and others merging into an elongated super bear, lol. Pretty cool in some ways, but I’m not looking for the overload of AI slop to come.

[@pcread]

AI videos are like dreams, the consistency of content is just not there yet.

[@JoePiotti]

Ask Sora to make a video of itself.

[@anthonylinus]

definitely not worth $200/mo

[@DreaMeRHoLic]

The real question is... at what point can I drop Baywatch episodes and playboy pictures from into AI to create VR..... Just for the lulz I've been thinking about making a real movie after prompting characters/actors, but this also just makes me wonder (just for the mess to show people what could happen so they become aware) to use the snipping tool to get friends instagram pictures and create a random movie about an event that never happened. The problem is that i'm aware that some people will awake and see it for what it is and others will ID me as the problem and think i spend months on "doing something on them". Over all i wish there would be a law to add information into the file that it is AI content and maybe what AI created it and maybe who is the account that told the AI to create it? I dont know what information you should add and what should be able to be backtracked, but I got a feeling the media will be full of AI stuff soon and it will become a big mess to find the needle in the haystack of crap content

[@Axiomatic75]

While some of these clips are quite impressive, most of them aren't quite there yet. Maybe throwing more compute at it will solve some or most of these problems, but for now it doesn't look good (mostly consistent) enough to be of real, practical use. But it'll only get better from here.

[@randomdude2582]

but how do i know what's fake?

[@Malins2000]

I wonder how Tencent HunYuan video fares against it. good part it is open source ;)

[@aliasgur3342]

I'm wondering how in say 5 year's time, with our eyeballs and brains saturated by 'fake' video - whether we'll just treat it the same meh way we treat CGI?

[@jamestaggert4361]

200 usd is expensive fun

[@halnineooo136]

In a few years cinema is going to merge with video games eventually. AI real time generated interactive VR experience bespoke for every user. You'll be the main character of your own new movie each time you pick your VR glasses. The movie morphs and changes course as you make choices and interact with it. The story, the characters, the scenery, the dialogues, everything written and generated by AI on the fly. Cinema industry jobs gone. All of them.

[@tomoki-v6o]

Sal Khan will be surprised of this.

[@tunestar]

I really couldn't care less about Sora.

[@stevecaya]

Nice video

[@nicomoller3823]

Sora Account creation is still shut down for me..

[@MilesBellas]

Origami into smoke = the best one?

[@erikjonromnes]

Kinda Sorta, a sorta, a sorta, t’ SORA, t’ SORA, t’ SORA, disorder, disorder, disorder, a sword a, a sword a, a sword a, s-word uh, s-word uh, s-word uh… I can keep going… lol

[@Intemkt1]

I wouldn't pay for that.

[@sp00l]

Sora needed to be released 6 months ago. It's old news now and not really that impressive anymore...

[@DynamicLights]

Nah they couldn't release it. Come oh show some respect to all the people that worked on this.

[@Lerppunen]

Lol.

[@TheSacredGoldsmith]

How can it be so ridiculously good while also being ridiculously bad at the same time. It would be so frustrating wasting money trying to get the bugs out for usable videos for things.

[@Gael_AG]

I don’t know if it’s the model or people having issues building interesting videos. Camera movement and color grading / lighting looks off . Although its Impressing I can only see this as b roll and concept art videos

[@EricJames429]

Your inquiry regarding the value of Sora, specifically the question “$200 worth it?”, is unfounded. Regular monthly users of Sora still enjoy complimentary (limited) access to the platform. Individuals who generate videos for a living, regardless of the reason, can ascertain whether the monthly fee of $200 is reasonable. Without reviewing your video, I would suggest that if you currently invest significant time manually producing content that generates revenue, you can certainly justify the expense of $200 per month, which would enable you to produce output more efficiently using Sora or any other video editing model.