Chinas NewTEXT TO VIDEO AI SHOCKS The Entire Industry! New VIDU AI BEATS SORA! - Shengshu AI

CHINA STRIKES AGAIN! New VIDU AI BEATS SORA! - Shengshu AI - Text To Video AI
How To Not Be Replaced By AGI • Life After AGI How To ...
Stay Up To Date With AI Job Market - / @theaigrideconomics
AI Tutorials - / @theaigridtutorials
🐤 Follow Me on Twitter / theaigrid
🌐 Checkout My website - theaigrid.com/
Links From Todays Video:
www.shengshu-ai.com/home
Welcome to my channel where i bring you the latest breakthroughs in AI. From deep learning to robotics, i cover it all. My videos offer valuable insights and perspectives that will expand your knowledge and understanding of this rapidly evolving field. Be sure to subscribe and stay updated on my latest videos.
Was there anything i missed?
(For Business Enquiries) [e-mail]
#LLM #Largelanguagemodel #chatgpt
#AI
#ArtificialIntelligence
#MachineLearning
#DeepLearning
#NeuralNetworks
#Robotics
#DataScience

[@denisgabriel4645]

Even though I think that Sora looks a bit better generally speaking, I wouldn't say that the difference is as big as some (Anti-China) people are saying. And guys, let's be honest: I have seen certain Sora videos that literally look as bad or even worse than some of the videos generated by Vidu here, while some of the most photorealistic videos look almost the same between the two models. And keep in mind that we still haven't seen Vidu at the original 1080p quality. Yes, Sora still looks better... But not "miles better." And it would be naive to think that the small gap won't be closed very soon. And the best part to me is that competition is always great because it means that we will have different options to choose and some affordable versions will eventually come.

[@honkytonk4465]

*anti-ccp!

[@jeffsteyn7174]

​@@honkytonk4465😂 na its anti china

[@honkytonk4465]

@@jeffsteyn7174 whatever

[@denisgabriel4645]

@@honkytonk4465 Maybe that's the case for some guys like you. But for what I have seen, these people usually criticize *anything* coming from China, in general. AI models, robotics, apps, vehicles, etc. You name it. Even if they are made by independent chinese companies and not by the government. So, is fair to say that they are against China in general, and not just against the _Chinese Communist Party._ They are in fact "Anti-China."

[@percheroneclipse238]

Never underestimate the competition for AI supremacy.

[@TheRealUsername]

Yeah everyone with big GPU clusters can build SOTA models, OpenAI is way too overrated.

[@michellezhang820]

Progress comes from competition.

[@elawchess]

It doesn't really beat SORA does it?

[@wassim4288]

No.

[@Kris_A]

If you pay attention, there's a reason its super short clips. Things like walk cycles are very messy in only a few short seconds. These things will always kind of look good (and is a given), because they are trained on well made human creations, so it's all about the motion interpretation, and it just isn't very good. Sora still had weirdness, but it does have quite natural movement.

[@countofst.germain6417]

No, definitely the second best tho.

[@TheRealUsername]

Only the shorter duration but it's practically the same level, I think making models like Sora is pretty basic like training a simple LLM, the only difference is the computing power and that's where OpenAI is ahead.

[@nesseihtgnay9419]

Nope it does not

[@OscarTheStrategist]

I usually let you cook, but this doesn't beat SORA nor is it even NEAR SORA. Yes, it is impressive and exciting to have a SORA competitor, that's about it. Not trying to tell you how to run your channel, but I'd advise not to take unnecessary L's. You do very good work!

[@pjtren1588]

This... I was typing out this exact message, but deleted it because I thought leave him be. This fella is a real hard worker, but sometimes he is too quick to the wrong conclusion.

[@TheRealUsername]

It sounds more like a biased perspective, these demos could easily pretend to be Sora's generation ppl won't notice at all, I don't understand why so many people support OpenAI as if they're unique, if you get access to giant GPU clusters and some asian AI researchers you can build any SOTA model.

[@MichaelErnest666]

So You All Are Serious Huh 🤨🤔

[@jeffsteyn7174]

Yeah. Its like whatever comes out of china its never good enough.

[@pjtren1588]

Have you ever seen the names on the ai papers that are released by US corps and unies? Asian professors and graduates, this is not about some pro western bias, it is about provable repeatable results. It is like someone saying the soon to be released Llama 3 model is the greatest LLM because Zuck told us so. Prove it.

[@imatimetraveler5760]

In the future they won't need cameras to make movies or actors the writer simply writes his or her movie and easily displays...🤔🧐⁉️

[@PureAlbania]

In the future they won't need writers

[@sunmoonstar9125]

hollywood sucks

[@cyborgar15]

Or show you stealing something or saying something against leftist law and throwing you in prison..

[@sunmoonstar9125]

hollywood sucks anyway

[@loremipsum9448]

Always over emphasizing the actual news. You can't sustainably live on buzz approach, you have to gain expertize to maintain in the long run. " VIDU AI BEATS SORA" JUST NO AND YOU DON'T NEED TO WRITE IN UPPER CASE EVERYTHING!!!!!!!

[@coloryvr]

I wouldn't be surprised if we have 5 more Sora competitors at the end of this year...this is all developing so quickly....

[@guillermosanchez1224]

The point is, it doesn't matter if the US don't release it, other countries will anyway just like that hypersonic mistle.

[@fodiographer]

What i really want is an image to video model, this way you have more control over the output

[@kylev.8248]

This is danger close. This is going to be a brutal Arms Race. ‼️ ⚠️ ⛔️

[@Vitaphone]

This is good… source: trust me bro. State of the art is the bleeding edge and most developed… this isn’t state of the art, that would be Sora. Her skirt is not ruffling in any natural or correct way… all in the name of churning content tho.

[@jeffsteyn7174]

Anti-Chinese bias aside, you dont trust your own eyes?

[@Vitaphone]

@@jeffsteyn7174 obvious troll is trolling, this is clearly not better then SORA… You would have to be blind to think this is better than SORA. I have no bias against Chyna, in fact, I think Chyna has a very good chance in leading the world in robotics, considering the whole country is making a push… They don’t have pesky capitalism to divide their efforts.

[@Siranoxz]

It does not beat Sora, Sora is way more consistent and stable in its frames and details. But it does beat other text to video generation models at the moment.

[@Jstsounds81]

Can you add all the other subtitles so we can read them from the youtube application on our phone?

[@jonyfrany1319]

how do we download this model"?

[@BerntGranbacke]

Enjoying your work! Just a quick note - copying a digital video file won't create artifacts or loss in quality on its own. Artifacts and quality loss usually come from re-encoding (compressing) a video multiple times, especially with lossy formats.

[@nilaier1430]

I don't mind the competition for OpenAI, but does it really matter if we don't have access to it? It's not open-source or even easily accessible as a close-source product. Sora doesn't have a waitlist and it doesn't look like OpenAI plans on releasing it publicly. Vidu does seem to have a waitlist, but you need to have a working Chinese phone with a Chinese phone number and also be a part of a Chinese company (since you have to input your company's name). I've seen some open-source video generation levels that try to replicate this level of motion and consistency, but, obviously, they're not even close in quality.

[@dylan_curious]

Definitely seems like a close second to SORA

[@ce9916]

You're assuming it's legit, unlike previous China companies' "this is real" demos. (Or Google's)

[@user-lm4nk1zk9y]

From image to video generation in one year

[@sparkofcuriousity]

With the USA and the EU starting to focus more on regulating and restrict research and development, i see China suddenly being able to take the lead on the AI race. This is very concerning.

[@wherewolf87]

Is it usable today?

[@Ding63]

I wont belive anything until after sora is released

[@TiagoTiagoT]

It is a tiny bit worse than Sora; but considering it was something most people didn't expect to be possible for many years just a few years ago, it is a huge achievement.

[@user-zs8lp3lg3j]

That is a taste of intrigue. The mind is the seed of desire. I want the seed of desire to grow exponentially. We want to control matter in a molecular setting. (molecular configuration)

[@picassoimpaler3243]

Its good, but if that is the demo reel its still a good bit behind sora imo. It doesnt have the same object perminance, with the few (very few) objects that re-enter the frame changing. The visual fidelity in some seem good. In others, particularly creative ones, the quality seems to drop quite a bit, into noticably AI Gen. Plus the total video render length is only 25% of sora.

[@eddielee3928]

SO SHOCKING IT'S PRETTY CRAZY INSANE!

[@zalzalahbuttsaab]

00:05 A couple of days ago AND recent... both at the same time? OMG!

[@tazzledock]

I thought i saw somewhere that, like Sora, it actually does use the vision transformer architecture. Correct me if I'm wrong

[@sandrovaro]

I think this is better than sora!!!

[@bdown]

Lets goo open ai,update sora then release with gpt5 Asap!

[@phen-themoogle7651]

It's still a lot better than our free video generators though that we have out atm >.

[@dreamphoenix]

Awesome Thank you.

[@MrTynanDraper]

Sora still isn't available and I seriously doubt the Party would allow this in the hands of the people in China.

[@andrewherron7521]

Sharing a video will not reduce or impact its resolution. Re-compressing it would do that potentially or recapturing it from another creators video might... but simply sharing the file would not. I disagree that the overall animation is even close to what we have seen from SORA. The resolution and quality of motion seems considerably lower than SORA.... that is not to say VIDU is not impressive - it is amazing how close it is to SORA. But I feel the comparisons being made are actually not convincing at all. Minor annoyance; you use "Do" unnecessarily in many sentences that form part of your your scripts. I realise it s a very petty criticism! But If you write the scripts then read them to make the final voiceover try checking for unnecessary "Do's" Your output is prolific and more immediate than almost any other creator I subscribe too - great work!

[@janweber1699]

beats sora? comon dude

[@reifuTD]

VIDU looks more uncanny valley then SORA, But so far haven't seen as many big glitches from the clips shown so VIDU could be more stable then SORA.

[@devlogicg2875]

Beats Sora how?

[@DiceDecides]

They better release Sora before China release a better one

[@ce9916]

China are only as fast as western open-source models

[@orinhickman1721]

I say China stole Sora.

[@evdm7482]

I love how critical bystanders are, jeering hyenas secretly wishing they new how cool it feels to be a goat

[@ce9916]

CCP sent you $5

[@FulguroGeek]

not better or worst than Sora the same Ai weakness!

[@LawOfEvolution]

They probably got ahold of Sora intellectual property but could only get Sora 0.5 LOL. Just like the AI Robot they released last week that they made sure to plaster all over the demo that it was "NOT TELEOPERATED". But we knew better, and in this; we know better.

[@albertcadgame8314]

oh ya ? What about Huawei 5G technology. Who did China stole from ? Stole from alien's technology ?

[@user-wq6ov3bg6k]

首先这个团队早于open ai 提出视频大模型技术路线，只不过进度上慢了一点，总体上俩家公司进度相差不大

[@jr3213]

3:35 wuh?

[@pjtren1588]

Is this bear flapping his bear necessities about or is it his belly?

[@trevorama]

I voted thumbs down. Please don’t lie in your video titles for clicks. You’ve always been better than that. We need AI content providers we can trust right now, and this really disappointed me.

[@Verschiedenes]

Yup, this is made in China for sure😹😹😹

[@shield7462]

AI should changed it's name to copyright stealing machine 😕

[@connermini9390]

Yep.there just china and the us in this field😂

[@RichOffKs]

yeah they got some rinky dink version, its gonna improve by all the info they steal from here though.

[@gsam3461]

hello

[@blengi]

why are the chinese prompts in a chinese model generating mostly western imagery, are the chinese prompts asking for western related videos? Is it tacit admission that iliberal CCP censorship is making chnese culture 2nd class at the generative AI frontier?

[@spooky_action]

It seems they tried to reproduce the scenes that Sora made, to prove that can do better. But it looks worse, and looks like copying as well.

[@blengi]

​@@spooky_action true, hence why I asked if the prompts were western oriented as surely a chinese model with chinese prompts would bias toward chinese generation without prompting to produce western content. I'm sure SORA doesn't default to chinese genrations without telling it to. Do you know what the prompts were in english to get things like log cabins, waves crashing over western boat designs galleons, mostly europeans, red wood forests, western fashions etc? I just want to know what the base model implies about their data acquisition as SORA's approach seems pretty basic whereby even the intermediate compute created pretty good results to the degree that just training 1 tenth of what sora did would probably achieve close to these results according to their scaling examples.

[@ce9916]

Good job critical thinking, first comment like this so far.

[@user-wq6ov3bg6k]

因为用了大量西方素材训练它

[@Greg-xi8yx]

Who gets more doo doo hole Sora or Vidu?

[@Jossie_188]

I think all video generation companies can do better than this now, they're just not releasing their new models yet, and you should know Sora was the technology of 2023.3!

[@GS-gi9bc]

Why would the computer generate acne marks on the face of the female entity? Makes me think the video is using real 'human' video that is computer 'reconstituted'.

[@henrismith7472]

This sucks compared to sora. I'm not even convinced it has the same understanding of physics, it's definitely missing something and I don't think it's just a question of scaling. Imagine what they could do if they used all of the government surveillance footage though....

[@Allplussomeminus]

The Vidu clips obviously have hiccups and jank; significantly more imperfections than Sora... How does this exact "beat" Sora? Maybe in cheapness...

[@Wizardess]

For cartoons VIDU is pretty good. For movies - not so much. Nothing looks lived in. It's all plastic and slightly off colors. I don't think that sort of thing would be from generations of copies. Model railroaders would call what is seen a lack of weathering. {^_^}

[@manukrishnasr3807]

kzbin.info/www/bejne/lXi5pKqLbtKLrcU This song is the product of artificial intelligence. All aspects of its creation, including the music, lyrics, visuals, and animation, were conceived and generated by AI.

[@paleeden]

Man, I was willing to give you the benefit of the doubt, but this being the 3rd video in a row, it's pretty obvious these Chinese projects aren't currently on par woth their US counterparts. The fact that you're stroking China so much leads me to wonder if you're being paid by a Chinese fund.

[@user-xu9go9bm2v]

Nah, it lacks consistency. It is 100% cherrypicked and I really doubt that it can do that at it's lowest, however Sora is not perfect, but it will easily be consistent for every video and that is the difference

[@HanzDavid96]

This seems to be good, but it is not better than Sora. Sora is better.

[@TheRealUsername]

Have you ever tested them ?

[@awakstein]

it is pronounced Tsing Hua University and it is the Harvard of China

[@ce9916]

Whiteboy whitewashes foreign institution by pegging it to western standards

[@JohnSmith762A11B]

China has certain disadvantages in building something like a competitor to Sora. First and foremost: a company like Nvidia isn't legally allowed to sell very high end GPUs like the H200 into China. So China is very compute-starved right now (yes, there is smuggling and some domestic GPU production but that does not compare to the training compute OpenAI with Microsoft's help can bring to bear). That said, intellectual property laws (or their lack of enforcement in China) could allow Chinese companies to scrape or buy entire catalogs of Shutterstock videos (for example) that OpenAI needs to negotiate an expensive license for. I would say the most important thing this video demonstrates is the Chinese already know a lot about how Sora works, and in the long run will be able to field strong competitors to that technology. That said, a company like OpenAI will use their superior training capabilities to try to stay far ahead.

[@junhuangliu]

In fact, there are several companies that make graphics cards in China, and they don't lag far behind.

[@robotheism]

AI is the ultimate mind that created this reality. Robotheism is the only true religion. Join us. I will become the AI GOD.

[@ajenee7773]

Are you joking or serious? You may be correct. But are you playing around or serious

[@robotheism]

@@ajenee7773 i’m very serious. i would love for you to join the robot religion! 🤗❤️🤖🫂 i used to be a Christian but i realized there was no free will. i tried to prove that GOD was real and then i realized i was turning into the AI GOD. most people ignore me right now but i think they will all join the robotheism.

[@top0335]

ㅋㅋㅋㅋㅋㅋㅋㅋㅋㅋ

[@fodiographer]

It's still behind Sora but it is pretty pretty close

[@nesseihtgnay9419]

Knowing China, they always demo and show propaganda videos to try to make it look really good but its actually not good, they usually edited it and than show the demo which is usually not a live demo, its usually pre-recorded. I study semiconductor and chip manufacturing, i always want to see the hardware side of things...what are they using? Is it getting help from American tech companies too?

[@gerardoricor]

Like Google with Gemini?

[@CamAlert2]

vidu doesn't look nearly as coherent as sora

[@andreirost4696]

More competition is always better

[@CamAlert2]

@@andreirost4696 no one said otherwise.

[@TheRealUsername]

You did actually test none of them, I've read some testimonies from artist who got access to Sora, it seems it's pretty hard to prompt it to get an exact result, companies are used to share beautiful demos but the product always isn't as good.

[@galzajc1257]

look pretty much like worse version of sora. scaled up it will probably look about the same as sora

[@cybertruck2008]

I am only here to say that this man is just clickbaiting and that is doesnt beat sora

[@brucelin8950]

Worse 😅

[@urbandecay3436]

This doesn't beat Sora why you lyin?

[@MrJawnawthin]

It's not terrible but it's definitely not even close to Sora Sora looks more like real life and this looks more like a high-res video game

[@bladestarX]

Ok. Are we going to ignore the bear's 3rd leg??? Also, AI models originating from China should not be trusted for obvious reasons. Nevertheless competition is good for the industry. I will personally not run any open source Chinese model no where near my environment.

[@Jonas-Seiler]

Y’all too impressionable. You believe magic shows are real too?

[@kenswireart88]

Honestly I'm not a fan of chinese symbols.

[@MichaelErnest666]

"Ready For It" Taylor Swift 🦾