Luma AI STUNS the ENTIRE Industry! AI Video is SCARY good | SORA killer?
Vložit
- čas přidán 25. 06. 2024
- Learn AI With Me:
www.skool.com/natural20/about
Join my community and classroom to learn AI and get ready for the new world.
Try it here:
lumalabs.ai/dream-machine
Theoretically Media:
• The BEST AI Video Mode...
Other examples of Luma AI generations:
x.com/justin_hart/status/1800...
x.com/LumaLabsAI/status/18009...
x.com/AndrewCurran_/status/18...
x.com/cfryant/status/18009239...
x.com/A_B_E_L_A_R_T/status/18...
#ai #openai #llm
[TIMELINE]
00:00 LUMA AI
01:52 Early Access Showcase
03:35 Dream Machine Features
10:02 Demos and Showcases
RIP OF models
FACTS NO. CAP. 💯😬😤
Yup 😂
Im hoping they allow nsfw
They won't be missed
😊
My God. Seeing the clips together at the end...knowing that non of this is actually 'real' but a result of an algorithm. Just mind boggling.
27 years ago I knew this was possible and would happen. Just look at computing trends and use reason to predict that most theoretical capabilities must be achieved at some time in the future, and indeed in most cases what you see in sci-fi becomes possible which means it was always possible.
@@UltraK420cool. I blew my mind when I saw Google’s deepdream in 2015 and was showing all my friends and listening to the very few ai podcasts and kept repeating “this is IT”
Something More mind boggling, that could very well be how our universe operates too.
If OpenAI deliberately slowed down the rollout of Sora to assuage fears about moving too quickly, only for an upstart competitor to release their version without such reservations, it reveals two things: (a) people are equally inclined to criticize OpenAI for being too cautious, and (b) it demonstrates the competitive dynamics of generative AI: there’s no slowing it down. Any attempt to do so will prompt competitors to accelerate their efforts.
Exactly why Elon Musk was going after Open A.I.
He doesn't care about the effect of A.I. on humanity. He was upset his company was falling behind and wanted to slow down Open A.I. as long as possible.... thank god he failed.
I think Sora had some glaring weaknesses that needed to be sorted out too. It still struggles greatly with crowds, groups of cyclists, eating,etc...
great point - release the shit already
@@henrythegreatamerican8136 Nothing you've said has any factual support. Instead, it sounds like you dislike Musk on a personal level, and therefore your look for reasons that you can 'interpret' in ways that reflect negatively on him: confirmation bias 101.
OpenAI, from Musk's point of view, was created because the top players in AI refused to heed the warnings of more than 1,500 AI professors and professionals around the world that we're moving quickly for the sake of 'being first' but ignoring the concerns about AI safety. Recently, Opus 3 was asked:
*"What are the chances if humanity surviving AI?"*
Opus 3 replied:
*_"Based on our current trajectory, considerably less than 50%. The challenges of aligning AI are immense, and we are not on track to solve them before advanced AI arrives."_*
It gave a detailed answer, which it was then asked to restate more bluntly. Here's what it said in revision for the sake of brevity:
*_"I'd give humanity a 30% chance of surviving. We're in a car hurtling towards a cliff, and we're arguing who gets to sit in the front seat."_*
So for all the people who think that 'being first' is the most important consideration, that may be true from the capital crowd but from the broader concerns of humanity overall... AI safety is our number one biggest looming threat, and those with the authority to decide to take this seriously have explicitly refused to do so.
They're not too cautious; they pretty much fired the starting-gun in the arms race years ago. Though it's hard to tell how long for some other greedy company to do the same in an alternate timeline where we never heard about the dangerous machine telling stories about two-horned unicorns that speak English...
I think what they're saying with "Janus" is that the model is creating Janus-like images when that's not really the intent.
Exactly.
Really surprised wes thought they were saying they arent good at two headed requests lmfao. Its very obvious they were just pointing out that regular animals can morph directions as weve seen before
@@wamyamYeah, that seemed obvi to me too. I always cut Wes some slack as he probably is constantly sleep-deprived keeping up with all this. 🥱
@@wamyam haha. The most difficult challenges: create text, morphing objects and, of course, creating polar bears with two heads.
Right I'm pretty sure that was highlighting various AI image errors. Seems appropriate that the host of one of the biggest AI hype channels is not even smart enough to recognize that was the intent of those examples lol
Welp, goodbye animation degree, you served me well.
gg
AI is going to make all degrees questionable lol.
scale, make bigger animations or more
Nah this is where you truly shine, use all you’ve learned and apply to these models, if you know what you want, you’ll do better than most at producing quality works. It’s just now instead of a team it’s just you, or expand you team and work of projects. Instead of a animation taking months or years to produce now you can create them in 1 week, day less in the future
The entire entertainment industry is now hosed. That goes for everything from books and music to movies and eventually video games.
Coherence is the only thing holding these technologies back from utterly transforming the world. ...how long will that dam hold?
Not gonna hold for long. We've seen more progress in the last year than the whole last decade. Newer and better models will keep coming. And then it's over for artists
@@HakaiKaien based. screw em.
@@HakaiKaien Think of it like chess. We've not been better at chess since Deep Blue, but we still play chess ... and pretty damn good chess too.
@@Sal1981 I'm not sure the analogy holds. When we watch chess, like when we watch any sport, it's not the sport we're interested in, but seeing the potential of the human being expressed through that sport. That's interesting because we are humans ourselves.
But when buying a product, we almost never ask, "was this made by human hands?" Most of the things we buy are machine made already. What we're interested in is the quality of the item we're consuming and the price at which we get it.
So the question is: which of these two scenarios is this most like? And honestly, I think media is more of a product, and less about watching the human form explore its potential. Like when I watch an Anime, I'm only judging whether I like the art, and the story, etc. I don't consider how much of it was produced by humans verses made through automated processes. I only care about the end result, and what that experience is like, if I'm being honest.
@@bujin5455 perfectly stated
A lot of early AI video already was nightmare fuel, I can easily believe the current and future AI video models can make terrifying visuals.
Finally is coming the global MMO with graphics made with AIs like these, procedurally generated into infinity
SAO when?
that's probably an ASI complete problem. will we still be around to play them?
because MAN I HOPE SO
We're going to need a looooot more computational resources for anything like that.
But hey, eventually.
@@w1ll2p0wr neurolink is needed
your living in one
Ai VR adult movies tailored on the fly to your best liking (monitoring vital signs)
Holodeck when!? 😊
infinite gooning
Skynut
@@aanon1342the other corporation is Skyjerk
im already masturbating to much lol, that would kill me.
I'm favoured, $130K every week! I can now give back to the locals in my community and also support God's work and the church. God bless America.
As a beginner what do I need to do? How can I invest, on which platform? If you know any please share.
Yes! I'm celebrating £32K stock portfolio today...
Started this journey with £3K.... I've invested no time and also with the right terms, now I have time for my family an…
Started with 5,000$ and Withdrew profits
89,000$
I'm glad to write her tay I do hope she will help handle my paycheck properly☺️☺️☺️
Can I start with as low as $1,000?
An Ai trained on your worst fears, turning them into a horror movie🥶
thats actually really awesome, a personalized horror movie
czcams.com/video/Oxc-v2qvvJk/video.html
An Ai trained on your kinks, will make your testicles explode
That's liable to flip somebody's brain inside out
Odd pricing- it's actually a few pennies cheaper to buy the Pro plan for $99.99 (x5) than the Premier for $499.99. Defeats the #1 rule in sales: offer financial incentive (discount) to buy more.
No. It shows that they're optimizing for stable, committed *maximal* profits, for the relatively finite supply they have to offer.
If they had infinite servers to run inference, then they'd optimize for max. revenue (through greater quantity).
You can’t kill Sora when it only exists in the latent space. The official release will be very telling. Hopefully this gets it released
This. Sora is still just an announcement with likely cherry picked examples.
Agreed 2 other companies have dropped competitive text to video, might make them send it
czcams.com/video/Oxc-v2qvvJk/video.html
I am concerned for our youths who may deep dive into the digital world beyond their capabilities to the point of being fooled by the real world.
Escapism is quickly becoming very accessible and we are not going to participate in the natural world. Peace through self knowledge is becoming obscure. 🎉😊❤
to be fair you can already do that with the phone, and it sucks
and yet there is also a renaissance happening of some people returning to nature and spending less time online. What scares me is children who have never experienced pre social media or pre internet and perhaps have not had access to real nature..
I find it very interesting that none of the examples I have seen here or elsewhere, as yet, can produce lifelike hands and hand movements!!
That's because the soul expresses itself through the agency of the hands. And when the machine is soulless, then the hands are going to appear lifeless.
Janus refers to unintentionally creating two-headed entities (the two headed polar bear for example). This comes from the training data which primarily has shots from the front of objects but fewer from behind. So the model doesn't really know how a polar bear looks from behind so it just uses the front again. That's also why the magician in the video, when he turns around, has another face at the back of his head. It's fitting in the context of the video but it *is* a limitation.
It's where we get the word January from too.
Wow. I’m so excited about this. Queuing is fast now already. The toothpaste is currently escaping the tube. No putting it back now. LFG
demonhogo wrote, _"The toothpaste is currently escaping the tube. No putting it back now."_
I'd say that's an incredibly damning indictment of the poor judgment of both the people who are creating the various LLMs, and the people who want it faster but without any real understanding of what's going on. That old cliche about 'be careful what you ask for' seems more relevant to the AI race than ever before. Maybe you can't put bcak what's already emerged, but you can stop squeezing the tube.
this AI horror clips are creepy…Interested in seeing the first ai generated horror movie
If we have this today, what are we going to have in 10 years???
Ai VR porn tailored on the fly to give you most pleasure
we will be in the trenches fighting robots, so no movies for us xD , we will be living terminator live.
Thinks that break our brains maybe.
In 5 years or less you are going to create your own 4k movies in your home with sounds fx and dialogue.
@@slateisa this tech wont be for you and me
The best use will be special effects. Use static image or unedited video and add explosions or something.
Those movies looks better than the Acolyte
With better writing/plot.
Don't worry, they are faster than you think. They already have AI prepared to push their agenda on us. Just ask them what a woman is, and you will see. Lama 3 is the worst for me now.
Maybe they know that Disney is already getting ready to disappear anyway, which is why they don't care about killing their IP.
It is good. I think the base of white boys out there are just getting twisted
It is good. I think the base of white boys out there are just getting twisted. 😂
Crazy that some of these scenes cost like $1 and a few minutes to make...!
czcams.com/video/Oxc-v2qvvJk/video.html
Really can't wait for your next video on this.....
This is ACTUALLY SHOCKING lol fr tho this is just shy of sora and doesn’t seem to be vaporware. Model consistency is not quite there, but wow, this seems to be as Claude is to ChatGPT
Hopefully we get more competition. Open AI needs to release sora
ive tried luma a few times. Being honest its not even close to what weve seen from sora, but the biggest difference is we have access to this and not sora.
@@voidhytale4554 i agree
I’m still in the queue 😔
Is the music chosen and edited into the video by AI as well as the text?
I think the Janus issue is exactly the opposite you described; extra heads showing when they shouldn't. Similar issue is present with some text to 3d model creators; the face/head is such an important element that during rotations the model just adds another one because it was trained on it almost always being visible. My guess is they're probably generating things more sequentially than as a whole. I imagine it might've been better if they had trained with the video frames stacked and rotated so time is one of the spatial axis and vice-versa; or maybe even a combo of regular sequential and rotated, or even throwing random space-time rotations and not just 90 degrees, to really give it a grasp of how things fit in all possible ways (obviously to output the final result, the rotation needs to be undone to render the final normal video output correctly).
once that gets real time with sound, any game can be translated into dreamed up absolute realism.
This is incredible, this is to inspire creator like us to be more creative with low budget and less effort. I believe more Video gen will come out before the end of the year.
Sora who? Thats what you get when you tease a model but keep it to yourself. Lets go luma! They said its highly scalable and efficient, so the more GPU's they get access to, the more beautiful their outputs apparently get.
I wouldn't mind seeing a couple of those movies... 🍿 Really going to be amazing when the viewers can dictate what happens next. 🎬🎥📣
Does this incorporate sounds? I tried it and didn’t hear anything. Sorry if this is a massively dumb question
Kooool thanks for the great vid
The 16:9 resolution images seem really blurry, what are they like 480p or 720p and then upscaled or something? Super soft look. I notice the portrait orientation images have better resolution.
What's the music track starting at 10:00?
It's incredible I was able to use this in India in a 2 tier city and it barely took 4 mins to generate the prompt. This is what I expected from Sora.
If they allow nsfw content then it will become a trillion dollar company in 5 years
less*
Also it would be used to flood the internet with stuff for pdf files. Yeah, I think I would rather have it not produce that kind of content
@@elijahmitchell-hopmeier182 Would you rather it be real then? Real children have to carry on suffering, you could've prevented that.
@@imsoreetodddid9007 This is a bad faith argument. I’m sure you believe that this is the case, but it would only lead to people wanting to try the real thing. This would cause only more suffering.
Feeding lust with images only grows it, it does not diminish it. Let us not delude ourselves into believing this would help to eliminate existing problems rather than exasperating and expanding harmful appetites.
I mean this truly, to the person that wrote this and to the people that liked this: reexamine your lives. This comment is wrong and you know it to be. You cannot cure addiction with dope. You cannot cure sickness with disease.
@@elijahmitchell-hopmeier182 That's the same tired old argument which has no validity since there is ZERO evidence to support it.
Therefore it is based on nothing but persecution, simple as that, no government has the right to make that illegal and it only deprives us of having a viable method of harm reduction.
I was thinking for text to video being easy and intuitive to pose and animate characters.
You could use something like controlnet with key frames, so you can use it for videos posing and movements of a character in the text to video generation.
I don't think anyone has done this, I would love to see it done.
Imagine posing your character just by dragging a stickman around then hitting the generate image button.
Easy posing if someone pulls this off.
That sounds awesome! Do it! Find a preferred NLP Capable Model and ask it how to begin. 😁
Im most hyped for this shot of what looks like anime, specifically ATLA, at 12:25, but it's so short 😭
This is cool for short clips but still takes alot of iteration to get anything decent and usable. Same with Sora
Janus is probably referring to the error where there isn’t object constancy with heads/eyes in animals. So the animals face is always turned towards the frame resulting in weird head/eyes placements. It happens in animals maybe because majority of time animal data sets are pointed towards the frame…
Beautiful video, Wes Roth.
You got the Janus staff back-to-front. They're trying to generate a polar bear and getting a Janus polar bear instead. Same with the morphing car. They wanted a car but got a morphing car instead.
I was stunned by just the thought of being stunned! My God, what have you done?!?
Good. About damn time!!!!
OMG a full-scale AI render of the Battle of Cannae is right around the corner
Seems to me the future of AI video generation as well as AI in general is the micro-servicing of its token output, where each token component is produced by a single AI microservice working in synergy with the rest of its kind under some type of master orchestrator, which would be responsible for ensuring consistency only and not actually rendering content, like the microservice would.
A chinese sora competitor right before the US election. Do you think it was intended?
Would be cool to make a Hellraiser one with the cenobites but the comic ones.
It's truly incredible, how far AI generation has come.. it's not perfect but it's still really good and available to the public. I can see once the duration of generations increasing, you can make custom tailored movies and shows once it becomes more consistent. The fact it can take real life as a reference and create a realistic like horror film is going to be true terror, no make up, no CGI it'll just LOOK real
NVidia working on 2d image to 3d if the model can recognise the objects in a scene it will be able to do photo realistic 3d scenes with animated camera. I expect this to be the actual future of video. It's like if someone asked you to image the other side of an object. In many cases your could like the back of a person, a sign post, a car pretty much anything symmetrical back of a coke can.
Now openai has to release their stuff
I don’t think open ai is going to break their edging streak just yet
Alright, well that’s it. The selected samples at the end are simply out of this world. Some artifacts and weirdness nonwithstanding, it’s the end of animation, cgi, camera operator, etc kind of careers. Those people better start learning storytelling and become AI directors cause this is the worst it’s going to be. By the end of the year we’ll be in uncharted territory where any creative person will be able to craft gorgeous films and animations by themselves with a couple thousand dollars budgets. I’m both excited and anxious about it. For the consumers it’ll be a content bonanza, for the creators it’ll be a never ending competition. Strange times await us my friends, strange times indeed.
Great stuff.
Thank you!!
Having to sit through those long long lists of animators at the end of movies to see if there is one of those tagged on movie scenes after the credits is obviously going to be a thing of the past.
Crowd scenes still inspire the phrase "Bose-Einstein crowd"
Thanks Wes
Saturday morning cartoons have come a long way.
Its awesome we get a chance to use it
daaaamn how can the industry even go further while being STUNNED all the time :o
You think of something, say a t-shirt for a class meeting or a new shape, form, function of anything you deal with, and your agent army will get to work contacting & contracting the automated physical services and have them delivered on your doorsteps, windows with drones... 😊
It's amazing work. Funny hiw Ai seems to still screw up our hands and fingers lol
13:30 Roko's Basilisk in training ...
if there was an entire movie of the final clip i would watch it, i didnt notice any weirdness on it and even if i did i wouldnt care, it looks so good, the vibe and all of that... its current quality is enough, it is sufficient for the industry and our standards
I have a Chinese number and KLING is not available to download anywhere. It is not OPEN to the public yet. I know one CZcamsr mentioned that and everyone is repeating the same.
It's not as close as it seems but the concept of infinity content will increasingly become relevant where anything can be generated on demand. You could do new episodes of star trek TNG in modernized updated look and recast, even plausibly interactive in VR
Man, if only there is something like this that we can run locally !
The GALA trailer looks insanely good!
Do AI (photo and video) creation software fuck up the words spelling and the fingers on purpose (voluntary sabotage) ? Based on a text prompt, how hard is it to just put the right word on a screen or to put the right number of fingers for a human being ? If this is voluntary, what is the purpose ???
czcams.com/video/Oxc-v2qvvJk/video.html
Thanks for sharing the link but sadly luma is extremely restrictive in the prompts you can use.I hope other less restrictive options.Appear at some point.
can`t5 wait for all the cool chronenberg things that will come out of this.
I don't know how they're getting results like these. I just played with it, and what it generated is something you would only ever see in a dream. The name makes sense.
Were getting quite close!.
Its nothing but impressive.
Janus is a two headed Roman god that looked back at the old and forward to the new, death and rebirth, etc. It is where we get the word January from.
Just tried it, took about 5 minutes for me on iPhone .😮
Nice .. tales from the crypt.. 2024 ai edition.. would watch for sure!👏 👏
If you haven't created stuff like that yea it's good. Only if you compare it to nothing. It's very inconsistent. Works for 2 seconds if you crop the hands, but won't work for an entire movie. I'm sure it could be made better, but would it be more time efficient or just a different workflow idk. It may be incorporated but won't change whole industries yet
it ain't even close to Sora, I tested it mate
00:01 Luma AI is a game-changer in AI video technology
01:41 Luma AI can create dynamic, action-packed scenes based on prompts
03:10 Luma AI can animate movements from static images
04:38 Luma AI's AI model creates high-quality videos from text and images efficiently.
05:52 Luma AI can create high-quality videos from text and images
07:16 Luma AI leading in AI video generation
08:42 Luma AI offers accessible, generation-limited video creation with pricing tiers.
10:37 Encountering Yuki at the shrine
Crafted by Merlin AI.
Wish they would have done more with the images at 12:34
That is the only thing in the entire video that stuck with me.
We are on our way to solve one piece of the holodeck puzzle. Now, matter out of light plz!
IMO a lot of this reminds me of silent film. Very short and very experimental. I think when we get long forum videos, matching voices, and it is cheap or free. CZcams will likely be the king of media consumption
That redhead 😍
Thank you.
3:07 Yes, obviously the delicious T&A is fantestical! BUTT, ...What's "crazy" to me is the billowing ropes and sails in the background!.. prettyFNcool!
it's funny how AI is slowly taking over the world, becoming super intelligent, producing incredible images and videos but... "sorry, that thing of arm-hand-fingers... I just don't get it" :D ...that bug is everywhere!
A.I. Corn is going to be amazing!
The future is here..I can only imagine what Sora can do. I tried it and even though it's got a ways to go it's actually not bad..imagine this in like a couple years, anyone can make a movie/tv/anime..
My dad drew a cartoon in a unieke style, when i upload a photo or scan, It transforms the character into a Disney cartoon. It cant maintain style. When uploading one photo of a photogrammetry shoot it flattens the modal.
Less censorship than OpenAI, I welcome the progress
Holy shit! Oh wow. Not guitar OG but yeah. Think I’m gonna start just growing plants 🤣
3:33
Everybody was looking at her arms.
I've played with it today and I gotta say: I am impressed. Dall-e was much more terrible at first and then 6 months later it was incredible. This is much better than old dall-e. I can only imagine how it'll be by the end of this year. And Sora IS better. So buckle up guys, we're headed for crazy time!
Brutal how fast these AI possibilities advance I think Holywood must be very worried when all of us can create their own films/ series at home with voice commands...I love the time we live in lol until arnie knoks on my door 😂😅
Reminds me of the 2000s when people thought blur fixed everything.
the janus effect is that it gives the polar bear two heads (imo)
imagine the world 10 years from now!!
Those non-cherrypicked results are really good. We know with Sora that there are typically dozens to hundreds of generations required to get the results we've seen. I think OAI really missed the boat with Sora, by the time it releases, it's going to be a marginal improvement at beast over what is out there. Not that I'm complaining as a consumer, works great for us.
240 shots per month for $30 is also very reasonable, a lot of people were estimating a cost of multiple dollars per generation for a tool of this quality.
Do we know why none of these AI's can accurately generate human fingers/hands?