6:44 your translation is correct, but only covers the first part before he pauses and speaks the second sentence at this timestamp. The second part translates to "if you haven't eaten apples for 1 year, then you are allowed to steal some apples!"
I can't run it locally on account of the horse power-- but, I have been thinking a good video might be a lesson on how to rent a GPU and run one of these models in the cloud?
This is good👌Mochi 1 originally required 384 GB of VRAM and FOUR H100s to run. The Hunyuan Video Model basically only requires 10% of that - and that's even without quantisation yet!
Thx Tim. Appreciate the open source heads-up. As you go through these products in the future can you briefly mention when they're restricted to paid plans?
1:47 I love that their motorcycle rider training footage includes dudes in g-strings LOL. Internet gonna internet. The opening sample (guy with rifle) had some amazing foliage. That is my dead giveaway... when leaves look more like popcorn than leaves, it's AI.
Haha, I was wondering about that motorcycle guy...I was like, is it? Yeah...you just confirmed it is! haha...well, not gonna judge, AI or Not, you live your best life, One-Wheel Motorcycle guy!
Hah, there's a reason I keep saying they're the "Crowd Favorite" Those guys bought ALL the good will. I haven't seen any other platform navigate the Free to Paid dance as well as they have. Kudos to Minimax!
HA! I was hours away to start learning Domo for fusing around with some anime ideas. Now you say MINIMAX my favorite video model of all has added these marvelous models? It really is Christmas. 😁
So close with the translation Tim! Tiny nuance towards the end: "I haven’t eaten apples for a whole year, so I went to the supermarket and stole a bag of apples. Do you think I should apologise? Just because you haven’t eaten apples for a year, does that mean you can steal them?" Testing the model requires chinese mobile number to register though 😔
yup! I've also heard it called the Zolly Shot. I think the correct name is the ContraZoom...but, who knows. I mean, if you call it the Vertigo shot or the Jaws shot, everyone knows what you're talking about. It's such a cool effect!
Very cool indeed, the basic trick it's nothing that complicated, it's just playing moving physically the camera forward while zooming out according to the speed of the device or viceversa, camera moving backwards while zooming in.
Have you tried to generate a 3d set using AI with just imagery from Google Earth Street view? Does it create 3d objects that can be imported in Unreal or Unity?
Yeeeah, I mean-- I think you'll see this model on a few platforms. Those who are using APIs to run Luma/Kling/Runway etc-- but what I think will be interesting is those who choose to actually do something with it, and tweak it to make it something new. That said, you CAN run this model by renting an A100 for like, 5 to 10 bucks an hour. It might be worth doing a video on that...I'll put that on the list.
Interesting how ai video generation behaves just like dreams. Things happening all of a sudden and changing to different locations sometimes without relation to one another.
Right. I've noticed this incredible similarity at the very beginning of the diffusion of generative AI art, now it's such a surprise for me to find out that it was not just my impression😅
Imagine being able to create ai environments in real time, as in you tell the ai what environment you'd like to walk through and it creates it while you are already walking through a generated environment...and that environment leads you into the next environment....long winded I know but super exciting stuff.
Being able to create scenes you can pan around and place characters in for scene purposes would be awesome I have to go check this out. I laugh every time Sora gets mentioned.
11:10 regarding the 3D shape of an object, the term to describe the surface and shape of a 3D object is "3D topology". As more and more AI dive into 3D content, you'll probably need to start using the relevant terminology too! 😁
The coherence wall is going to be a hard one to overcome, the image gens can use conceptualization of learned token-to-visual patterns - but visual narration requires an extra layer of reasoning that I imagine will be a tough nut to crack.
I think the answer lies somewhere in 3d models that are then skinned over via Video to Video. All kind of in the backend and hidden from the user. But, using 3d as a "skeleton" has always been my thought. But-- there are also much smarter people than me working on this!
@@TheoreticallyMedia Totally a theory but perhaps artificial intelligence itself needs to be harnessed as the bridge builder, comparing and contrasting video outputs and stitching them together in a prompted way, thanks!
Unfortunate how these local installs won’t run on my 2020 16” 20G Ram 512GB MacBook Pro, it’s even struggling w the new Topaz suite I just purchased, Video upscaler
oh, man-- I'm running into that with my m1 pro as well. I'm thinking it might be a good excuse to bump to the M4. I've been waffling on that a bit. That, or maxing out a m4 Mini Pro... Or, just finally biting the bullet and building a PC. So much of the local installs rely on Nvida cards. GPU hungry.
I’d say Tencent claims so, or at least in their wording “better or almost better” haha. I don’t think so, personally- BUT, this being open source, it could be tuned up by the community to become something really powerful. We’ll see what happens with it. Currently, we’re kind of in an age of embarrassing riches. So I’m just enjoying it all.
Minimum: The minimum GPU memory required is 60GB for 720px1280px129f and 45G for 544px960px129f. Recommended: We recommend using a GPU with 80GB of memory for better generation quality.
THANK YOU!! I do love those little bits of trivia. That information is now stored in my long term memory!! Appreciate it! Fingers crossed it comes up in the next Trivia Game Night! haha
@ My wife and I actually got to seem them in 2010 near Beijing in 2010. There are 8,000. I known I've seen a couple of movies where they've been brought to life with ill intent...
its weird how much these videos are starting to seem like dreams. Like they're images created based on what the universe looks like, but they don't have to follow any of the rules, so anything could happen.
we getting off the pay wheel!!!!!!!!!! runway has been so cocky, and the service has slipped so I left and boom this drops the next day!!!!!!!! I'm not being super fake positive I'm so glad to see things shift to us... i said it next year our agents will be running most of these tasks!!!!!!! Yesssssssssss its time for the ones who want wealth to build and put in an effort to hustle for yourself off your own mind and skills.
I've watched videos of people trying to explain it, but I still don't understand the open source BUSINESS model. Why give all that training away for free?
To be honest, sometimes they just do it in the name of advancing progress. I know that was a big factor with the release of LTX's model. Like, legit for the good of us all. Tencent? I'm not sure about...But, hey-- not going to look a gift horse in the mouth.
Technically Chrome is opensource too. Having good open source thing like that may destroy buisness models of your competitors. That is why Mera llama is opensource, drive competitors out of buisness then start charging. Its oversimplification but game is complex.
Considering those are their cherry-picked best examples, uuuuuuuuh, the game hasn't changed except in that it's open source. In fact, their best results are not anything to get excited about. That said, considering how much Runway and Kling... cost, improving open source options are the way to go. More power to them. Lastly, Runway for image to video is not the best. Kling easily trumps it, as does Hailuo. If someone else does video to video I will happily dump Runway and their new slower than molasses "explore" mode on the unlimited plan.
Hi can anyone tell me if there is an online community of people like me making ai videos using various tools. Something like on discord or something. Would really help to talk to others doing similar things we could learn a lot faster. Please anyone let me know thanks
Tim i live for the commentary I'm cry laughing lol AI it's just llike people too much on its mind lol like my clothes in laundry lol needs to go in dryer lol
Tencent? Some might argue THE Chinese Company...haha. They're pretty massive. Although, they've got their fingers everywhere. From owning Ubisoft (games) to Tencent Films-- I think they produced the last Wonder Woman movie (which, I won't comment on), and Kong: Skull Island, which was MUCH better than it deserved to be. Seriously, that movie rocks.
It's all incredible. You barely get used to one advancement before another arrives.
Good stuff as always, Tim.
I feel like we're running on the treadmill like that AI Generated lady!! Just trying to keep up!!
One day soon, a new opensource 2 hour video with special effects and better than Dolby sound and we all become Steven Spielberg, and no one will care.
Every time I hear "Sora" I just roll my eyes unless we're talking Kingdom Hearts
😂 ME TOO!!! Lol someone else finally said it
SAME LMAO, I was a huge Kingdom Hearts fans until KH3 😓
@@Biajuuu at least we still have KH2 🥲
@@AustinsEdits peak KH game
Lmao I did the exact same
6:44 your translation is correct, but only covers the first part before he pauses and speaks the second sentence at this timestamp. The second part translates to "if you haven't eaten apples for 1 year, then you are allowed to steal some apples!"
wow that was a huge update video, thank you for it !
Always a pleasure to watch your refreshing news with that extra spark of humor. 😄 "Do you think I should appologize?" 😅
Omg, nice, thank you for posting this video and the wonderful news, thank you
Thank you, Tim! Great video, trusting that you'll really test it out for us !!
I can't run it locally on account of the horse power-- but, I have been thinking a good video might be a lesson on how to rent a GPU and run one of these models in the cloud?
This is good👌Mochi 1 originally required 384 GB of VRAM and FOUR H100s to run.
The Hunyuan Video Model basically only requires 10% of that - and that's even without quantisation yet!
Tim, you are absolutely the best UA-camr on AI Video today! 👍👍⭐⭐
I signed up for WorldLabs. Can't wait to try out some old photos
This was great. Funny too. Thank you❤
5:54 oh man i almost actually died laughing
Haha, I kept thinking-- need a few more holes in that ghost costume, Charlie Brown!
I almost spit the stuff I was drinking
Thx Tim. Appreciate the open source heads-up.
As you go through these products in the future can you briefly mention when they're restricted to paid plans?
1:47 I love that their motorcycle rider training footage includes dudes in g-strings LOL. Internet gonna internet. The opening sample (guy with rifle) had some amazing foliage. That is my dead giveaway... when leaves look more like popcorn than leaves, it's AI.
Haha, I was wondering about that motorcycle guy...I was like, is it? Yeah...you just confirmed it is! haha...well, not gonna judge, AI or Not, you live your best life, One-Wheel Motorcycle guy!
Very Cool!
This one I'm perfectly happy taking any heat for the "Gamerchanger" thumbnail!
Minimax is the best I’ll be releasing a 30 minute Christmas movie
Hah, there's a reason I keep saying they're the "Crowd Favorite"
Those guys bought ALL the good will. I haven't seen any other platform navigate the Free to Paid dance as well as they have. Kudos to Minimax!
Wut in Sam hell?...thirty minutes?...how many 6 second videos is that😂
@@willD.A.Vkarter I have a lot of time on my hands and yes they were each 10-6 second clips
I have to see this sub'd. 300 generations at least.
@ sorry the channel is glory house studios that’s where I make all my movies
HA! I was hours away to start learning Domo for fusing around with some anime ideas. Now you say MINIMAX my favorite video model of all has added these marvelous models? It really is Christmas. 😁
dope vid as always thank you keep the good work up ..also i like the jokes
We're getting to the point where we're seeing one jaw-dropping announcement after another every week. That is only going to become more frequent.
So close with the translation Tim! Tiny nuance towards the end:
"I haven’t eaten apples for a whole year, so I went to the supermarket and stole a bag of apples. Do you think I should apologise? Just because you haven’t eaten apples for a year, does that mean you can steal them?"
Testing the model requires chinese mobile number to register though 😔
Funny video at 4:20. Motorbike turns into bicycle and then into a rocket.
When you spotted that head...OMG 🤣🤣!!
Some people still don't have 80GB of video memory? Wow, they sure take their time upgrading these days.
Haha, I gotta scroll around Ebay for a used A100.
This comment brought to you by Elon Musk and Bill Gates.
Kudos to you for being able to focus on her watch and fingers...
At 10:48 the "Jaw shot" it's technically called "vertigo effect" if I'm not mistaken
yup! I've also heard it called the Zolly Shot. I think the correct name is the ContraZoom...but, who knows. I mean, if you call it the Vertigo shot or the Jaws shot, everyone knows what you're talking about.
It's such a cool effect!
Very cool indeed, the basic trick it's nothing that complicated, it's just playing moving physically the camera forward while zooming out according to the speed of the device or viceversa, camera moving backwards while zooming in.
Have you tried to generate a 3d set using AI with just imagery from Google Earth Street view? Does it create 3d objects that can be imported in Unreal or Unity?
I haven't done it yet personally, but I have seen it done. Yeah-- wild.
@@TheoreticallyMedia Which AI company did it?
I do wish these companies would understand their audience. This models requires a $40,000
Yeeeah, I mean-- I think you'll see this model on a few platforms. Those who are using APIs to run Luma/Kling/Runway etc-- but what I think will be interesting is those who choose to actually do something with it, and tweak it to make it something new.
That said, you CAN run this model by renting an A100 for like, 5 to 10 bucks an hour. It might be worth doing a video on that...I'll put that on the list.
That would be a great video! $5 - 10 an hour could be great value depending on the speed of video output.
Interesting how ai video generation behaves just like dreams. Things happening all of a sudden and changing to different locations sometimes without relation to one another.
Right. I've noticed this incredible similarity at the very beginning of the diffusion of generative AI art, now it's such a surprise for me to find out that it was not just my impression😅
Great this will be available in six months to consumer level
OMG that head popping out of nowhere was kinda scary
😂
Imagine being able to create ai environments in real time, as in you tell the ai what environment you'd like to walk through and it creates it while you are already walking through a generated environment...and that environment leads you into the next environment....long winded I know but super exciting stuff.
"Do a barrel-roll". Nice to know the grey hairs aren't lying. XD
Being able to create scenes you can pan around and place characters in for scene purposes would be awesome I have to go check this out.
I laugh every time Sora gets mentioned.
Wow tim please do the new minimax anime video the sketch option looks amazing and can't wait for 3d world access
2:43 😮
Ltx studio model is lot smaller and available right now.😊
11:10 regarding the 3D shape of an object, the term to describe the surface and shape of a 3D object is "3D topology". As more and more AI dive into 3D content, you'll probably need to start using the relevant terminology too! 😁
Although, I work for an animation studio, and the artists here usually just call it "geo" (as in, geometry).
@@IceMetalPunk yeah, actually that is also another term that is being used to describe a 3d structure. different words for the same thing. lol
Someone is gonna have to think out of the box, The Coherence bridge will have to be solved, otherwise we maybe hitting the wall.
The coherence wall is going to be a hard one to overcome, the image gens can use conceptualization of learned token-to-visual patterns - but visual narration requires an extra layer of reasoning that I imagine will be a tough nut to crack.
I think the answer lies somewhere in 3d models that are then skinned over via Video to Video. All kind of in the backend and hidden from the user. But, using 3d as a "skeleton" has always been my thought.
But-- there are also much smarter people than me working on this!
@@TheoreticallyMedia Totally a theory but perhaps artificial intelligence itself needs to be harnessed as the bridge builder, comparing and contrasting video outputs and stitching them together in a prompted way, thanks!
4:14 cue Benny Hill
Unfortunate how these local installs won’t run on my 2020 16” 20G Ram 512GB MacBook Pro, it’s even struggling w the new Topaz suite I just purchased, Video upscaler
oh, man-- I'm running into that with my m1 pro as well. I'm thinking it might be a good excuse to bump to the M4. I've been waffling on that a bit. That, or maxing out a m4 Mini Pro...
Or, just finally biting the bullet and building a PC. So much of the local installs rely on Nvida cards. GPU hungry.
I am lost, so better then Kling and minimax?
I’d say Tencent claims so, or at least in their wording “better or almost better” haha.
I don’t think so, personally- BUT, this being open source, it could be tuned up by the community to become something really powerful. We’ll see what happens with it.
Currently, we’re kind of in an age of embarrassing riches. So I’m just enjoying it all.
is there an AI App/tool for story boarding and movie scripting?
It's a pity that registration is only possible with a Chinese number(
Not this one. This is just downloading the code from Github.
tencent is a chinese company. maybe it's the reason.
How to install?
AT least they didn't drop this on friday :)
Minimum: The minimum GPU memory required is 60GB for 720px1280px129f and 45G for 544px960px129f.
Recommended: We recommend using a GPU with 80GB of memory for better generation quality.
You leave ONE disembodied head in your room and suddenly people think you're a serial killer. 😉
Robot helmet 😂😂😂
nice but don't they do an open source 500b or 1 trillion parameters models like llama/gpt but just 4 video...??
Not yet. There's much less data in text than video, so it scales easier.
Hi Tim,
I offer this as information not criticism... since I know you've an eye for detail... They are terracotta warriors not stone... 😊
THANK YOU!! I do love those little bits of trivia. That information is now stored in my long term memory!! Appreciate it! Fingers crossed it comes up in the next Trivia Game Night! haha
@ My wife and I actually got to seem them in 2010 near Beijing in 2010. There are 8,000. I known I've seen a couple of movies where they've been brought to life with ill intent...
6:59 I am always scared with translation software. I am sure someone got a tattoo they were told means one thing only to mean "I smell like dog butt".
12 gb vram quantization model someone make it happen for hunyuan ai video model please 🙏🏻
its weird how much these videos are starting to seem like dreams. Like they're images created based on what the universe looks like, but they don't have to follow any of the rules, so anything could happen.
👋 Open source? 👀 👍 👍 👍 👍 👍
Open Source, but HEAVY. haha-- likely most of us won't be able to run it-- BUT...you COULD. And that's what's awesome.
@TheoreticallyMedia
I'm at 48GB GPU in my Workstation and if I can port the remainder over to my Workstation CPU....... I might be in business!
We're lagging behind China in AI generative video.
Excellent. I've ordered myself a compatible NVIDIA GPU. I would recommend the NVIDIA HGX A100 640GB GPU - It's a bargain at about £160,000
Was this a joke, or are you rich? 😅
By 2025 the mis of 2025 we will get movie making model
Sora seems....So, So, So Long. RIP💀OpenAI☠
we getting off the pay wheel!!!!!!!!!! runway has been so cocky, and the service has slipped so I left and boom this drops the next day!!!!!!!! I'm not being super fake positive I'm so glad to see things shift to us... i said it next year our agents will be running most of these tasks!!!!!!! Yesssssssssss its time for the ones who want wealth to build and put in an effort to hustle for yourself off your own mind and skills.
this is a chinese ai tool ! i know some similar function tools like this
I don't know why but the segment of the woman running made me giggle.
uh.. physics... :)
I've watched videos of people trying to explain it, but I still don't understand the open source BUSINESS model. Why give all that training away for free?
To be honest, sometimes they just do it in the name of advancing progress. I know that was a big factor with the release of LTX's model. Like, legit for the good of us all.
Tencent? I'm not sure about...But, hey-- not going to look a gift horse in the mouth.
To get people to use your standard
Technically Chrome is opensource too. Having good open source thing like that may destroy buisness models of your competitors.
That is why Mera llama is opensource, drive competitors out of buisness then start charging.
Its oversimplification but game is complex.
*Sigh* You sound like my mom. "If you made it, you have to charge for it!" It's not always about the money.
45GB of RAM excludes 99.999% of people. It is impressive, but 720p with that requirements is a little disappointing.
Considering those are their cherry-picked best examples, uuuuuuuuh, the game hasn't changed except in that it's open source. In fact, their best results are not anything to get excited about. That said, considering how much Runway and Kling... cost, improving open source options are the way to go. More power to them. Lastly, Runway for image to video is not the best. Kling easily trumps it, as does Hailuo. If someone else does video to video I will happily dump Runway and their new slower than molasses "explore" mode on the unlimited plan.
would like vid to vid - so i can have my crappy architectural animation look as good as with single images
She has a watch?
There are windows?
Hi can anyone tell me if there is an online community of people like me making ai videos using various tools. Something like on discord or something. Would really help to talk to others doing similar things we could learn a lot faster. Please anyone let me know thanks
Tim i live for the commentary I'm cry laughing lol AI it's just llike people too much on its mind lol like my clothes in laundry lol needs to go in dryer lol
We need chinese mobile number to use it .😢
Another Chinese company?
Tencent? Some might argue THE Chinese Company...haha. They're pretty massive. Although, they've got their fingers everywhere. From owning Ubisoft (games) to Tencent Films-- I think they produced the last Wonder Woman movie (which, I won't comment on), and Kong: Skull Island, which was MUCH better than it deserved to be. Seriously, that movie rocks.
hahahaha
world labs = video games sora = yawn
what about animatediff v4
I now feel bad for skipping physics class.