The BEST AI Video Generator you can use NOW!
Вставка
- Опубліковано 7 вер 2024
- Luma Dream Machine tutorial & review. Sora-level AI video generator that's available now. You can even make action scenes and anime!
#sora #aivideo #lumalabs #klingai #ainews #ai #agi #singularity
Thanks to our sponsor Abacus AI. Try their new ChatLLM platform here: chatllm.abacus....
Try Dream Machine:
lumalabs.ai/dr...
Newsletter: aisearch.subst...
Find AI tools & jobs: ai-search.io/
Donate: ko-fi.com/aise...
Here's my equipment, in case you're wondering:
GPU: RTX 4080 amzn.to/3OCOJ8e
Mouse/Keyboard: ALOGIC Echelon bit.ly/alogic-...
Mic: Shure SM7B amzn.to/3DErjt1
Audio interface: Scarlett Solo amzn.to/3qELMeu
CPU: i9 11900K amzn.to/3KmYs0b
Thanks to our sponsor Abacus AI. Try their new ChatLLM platform here: chatllm.abacus.ai/?token=aisearch
Congratulationsn on getting sponsors too! will click and check it out to help you with your affiliate link
Okay
24:42 "She does not want to talk to me"
Wow, it is so realistic.
it understands reality
Women hate men according to what I've been watching 😂😅😢
Do you guys think there will be a time when people will be fed up with AI all over the place and start seeking for things advertised as "100% created by humans, no AI involved!"... or nah?
I’m more concerned on everything being possibly fake on the internet, news, propaganda etc.
yes, this scenario would be quite likely. problem is, how could we tell if it was 100% created by humans?
It'll be a niche like "cage-free eggs" or DJs that still use vinyl.
yes,seeking humanity
Tangible things are making a comeback. Acoustic music will be popular. Can't fake analog natural talent.
It's funny how the ones it can't do well. It looks frustrated about it, lol. Even the car one it drives it off the cliff as if acknowledging it did a terrible job. 😂
that man really hated the noodles!
@@theAIsearch he just saw that one dark web video and doesn't trust any noodles anymore. 😂
Would be interesting to play video games in real time ai generation, with a general theme to follow
You my friend are an innovative man thinking forwards towards the future
great idea!
I know! I am so excited for that! it's so amazing the temporal consistency that new AI video generator models have compared to yester year, hopefully they'll continue to improve and require less processing time and power and can run locally because as you know once companies start to use it, it's another subscription service to render your games, a genius business model indeed
@@wesley6442 Possible but creating games is not creating video. There is interactivity and that is almost impossible to handle for the next 25 years.
25 years looks like a typo lol, you mean 2-5 years right?
*Humanity:* Let's make AI, Free and Open Source.
*Greedy Corporations:* Yes! What a great idea! Now pay us if you want to use it!
Corporations: sure, we'll make it free & open source, but you still need our cloud compute to run it!
@@theAIsearch Exactly, and it’s better that way if it’s gonna require 128 GB of VRAM to run it 😂
I've given it a go 4 times now! Some incredible results, and some very strange. Either way, this is a blast to mess around with and see how my prompt engineering holds up in a new generative environment. Thanks for the video!
🤣didnt know people actually called themselves "prompt engineers", especially after all the roasting the term has got for meaning "i write in english"
24:45 prompt: beautiful girl talking to me. generation: beautiful girl turning away, she doesn't want to talk to me. wow, that is the most realistic generation!
it understands reality!
I felt like a few of those generations were very uncanny, some of them really made me uncomfortable lol, it made me think of some of my creepiest fever dreams
lol
That Chinese New Year Dragon was some serious nightmare fuel.
Great review! Exciting to see the progress in AI video generation. Keep up the good work exploring and comparing different platforms.
Thanks!
I've tried it, so cool! But, I still waiting for Sora...
Still realistic enough!
nice! how long did a generation take?
@@theAIsearch Just 30 seconds or a minute. Not too long... :)
@@theAIsearch took me a full day 😂
@@HelamanGile high five!
Imagine if someone else remade this video and only changed enough words for it to be considered "original". That's AI art.
I imagine prompts will eventually lead to camera scan saves.. where you can import pictures for details.. voices, environment details etc
@16:15 this was also my first test. It didnt work for me until I gave it a reference image to start from. Very, very promising.
😃
So moving objects and actions are the hardest part of video generation.
yes, and this is the best one i've seen for action scenes, though it's still filled with flaws
It seems to have issues when objects interacts with eachother, like eating, hand crossing etc.
I’m really enjoying this model
😃
This Luma Dream Machine tutorial is a game-changer!
I've been playing around with it for hours and the results are amazing. Can't wait to see what crazy stuff people come up with.
#SmythOS #ai
Me at the beginning of the video: It's here, we're fucked!
Me at the end: Phew, not today!
The good result come from good training data.
When the result is bad, it means they didnt gather enough footage to generate the best result possible.
Therefore they are FORCING you to use google account instead of random email provider because through google, they can suck more data from you and from around you.
i want something this good but open source, i don't want a company limit my desires
it'll arrive in a few months
I couldn't agree more! I felt this way with Udio and Suno censoring every little bad word or hint at something bad, that and chat gpt are so PG and censored that I honestly cringe when I use them.. I want to create some music about destruction or even some kinky hip hop stuff, and I don't want to give any companies my money to do it, I want creative freedom and liberty not censorship up the wazoo it's annoying
@wesley6442 it costs time and money to create this stuff. If you want free and open source you have to wait till people make it out of the goodness of their heart. Or try and contribute to it yourself.
@@wesley6442there are a ton of free and open source chat AI and image generation tools. But no company holding your hand through setup.
@@wesley6442 ChatGPT can be circumvented. I have some very graphic scenes
I tried it it's incredible,!! The best ai vdo ao far in my opinion.
😃
I confirm the AI engine is one or two steps ahead any other Video generator: it can really animate a scene adding lots of details. It understand and can do a lot more stuff than any other, also. Not perfect, anyway... the pure "text to video" generator often gives wrong details and, in general, result is a bit blurred on most image. (the crispy examples on home page are "cherry picked", for sure!) The "image+text to video", instead is really... really amazing and will give super sharp and super consistent results. At the moment, it think is way the best solution to get top quality AI videos. expecially if you feed it with the right images.
As CONs, the service is in a "basic" version, no options, no consistent characters, no video continuation,no upscale... nothing else than the "generate your 5 sec video". Is quite expensive, also: 30 $/month is the cheaper option.
I wonder how these prompts would work out if you gave it a starting picture made by Midjourney or another AI image generator, then you'd gain so much more control over the outcome. And with starting off with a picture, you also get consistent characters over more video generations, which you can then stitch them together for a much longer video. Then this thing get super interesting!
its hit or miss. i've uploaded images from SD near the end of the video, but the character was not consistent
Quantum computation is gonna change everything once it becomes online, this tech and everything else is gonna jump thousands fold.
Can you please explain me how?
@@dhruv_gaur52458 he doesn't know he is just traying to sound smart
a large number of problems are solved faster with a classic computer. quantum computers are only good at solving specific, certain kinds of problems.
@@jussivalter agreed. If quantum computing can solve the specific problem on how to make classical computers even more efficient then I'm all for it.
@@dhruv_gaur52458 The promise of quantum computing is that it'll be able to compute everything or mostly everything all at once.
Classical computations runs line by line, it's just that it's extremely fast at doing so due to the advancements of material & computer sciences.
So a good analogy would be building a brick house. Our computers right now will be building brick by brick, layer by layer, structure by structure. Quantum computing however, will be building the entire brick house all at the same time,as long as there is a layout, energy and the proper environment to do so.
So to apply this here to this tech, the A.I. would have in theory, the ability to dream up prompt generations in seconds/instanteously and maybe even do it better with less or without noise and deformities.
The challenge currently with quantum tech is providing the proper environment for it to function, energy cost, material cost and the pioneering of the tech to be able to solve the same issues as classical computers. It is a specialized tool at the moment but hopefully not for long.
It gives amazing results, BUT I wish they simplified the design. We’d need an interface to do what we want to do, we don’t need some fancy design animation on the website which will make the website loading slower and makes an Annoying experience. But I have high expectations of this amazing tool and I hope they make it easier to load.
I hope soon animators will begin to use tools like this one for us to get more than one episode per week or more animations per year 😅
lol
You mean corporations? like the CEOs cause I can tell you this stuff does not benefit anyone in the creative grounds
@@Deviaox those tools benefits everyone, of course now the tools are not as good as we would want it to be but in a few years it'll make people's job easier and less demanding plus, we as consumers, we'll get more products, and more is translated as lower prices which also benefits consumers.
@@Deviaox and btw in a few years even I (I'm not an animator) or anyone will be able to make entire animations or cinematics. Of course the more you know the better the results with ai tools, so I'll never get the same quality as a professional that have knowledge and experience on this topic. 🙃
@@kuroallen6419 you do know corporations are starting to replace animators for AI generators? Literally multiple artist/animators are losing their jobs because companies are laying off artist and animators so they can spend less and earn more. Sony pictures and other huge entertainment companies announced that they will be implementing AI for their production and started laying off artist/firing them. It doesn’t benefit animators what so ever, you guys can enjoy AI and I can’t judge that, but we aren’t blind we know that animators are getting fired from their jobs, few results are: hollywood is starting to collapse, lawsuits are happening, disney is losing it and independent animation is putting entertainment studios to shame.
I specifically told Luma to make the characters dance, but it must have not understood. That said it can do non-human characters and world creation in both 2D and 3D animation.
lol tryed it 2x and it failed "woman tossing a football" had a girl walking backward with a football in hand, the 2nd one generated another girl but her hand was morphing and she still didn't throw the ball
oops. yeah AI still sucks with hands and fingers
It's really fun to play around with, but I still feel like it has a ways to go before it's practical.
for sure!
A million miles it lacks character constituency alot obviously we need to be able to in platform generate cast and there attire
As well as general prompt adherence
Kling is the best Ai image generator man!
I've just made 5 videos back to back using the last frame to extend the animation. It is Amazingly consistent. Final is 23 seconds long.
awesome!
The noodle guy be like R U Komedy me 😂
lol
You need a Google account to sign up, I did yesterday but the wait time is insane cuz of the amount of user, I did one video but gave up on the 2nd one which I'll try again today
try with a new account
@theAIsearch I keep getting a network error, not sure what that means but would a new account fix this
Good luck UA-cam! It was nice knowin' ya!
tried it today it's pretty cool, not perfect but at least the public finally has access to it. I can't wait until it improves and also something like this becomes available to install locally because it sucks being locked in behind a credit system and the pricing on this thing is insane..
yes, finally! i hope open source will follow shortly
5:42 The guitar fret board is wrong though. Any guitarist would spot that they are not evenly spaced. The two in the middle are too close together
After few months YT & insta will be filled with such funny AI videos with crazy weird motion. That man eating noodles was hilarious. 😁
24:35 crazy, it's so realistic from an AI perspective, a girl like that really wouldn't talk to AI nerds like us
haha, so it understands reality
The food part is hysterical.
It is the best we can use now, and that is what matters. What is best among the ones we cant use, is irrelevant. Plus they have probably cherry-picked a lot. But can not see anyone making a good full length movie based on this, not even a good short p**n clip (since it has issues with eating), but still impressing and a step forward. Maybe it will take 2-3 yrs more before we get something that will really be a game changer. But step by step we are getting there.
That fight scene though. This it it guys. Start generating your Woo Ping movies.
yeah that was impressive
In a fight, someone always gets a Woo Ping!
Hmm I learned A.I loves offroading, wonder it does if ask to make a funny scene etc
5:27 the guitar's headstock is backwards. looks like locking tuners lol
Sora film took more than 6 hours to compute... on supercomputer! C'mon! fast generated and decent video is STILL OUT OF REACH OF ORDINARY PEOPLE!
bruh, i regret generating a devil fighting an angel and god watching and crying
Good said- INSANE! :D
And it's just the beginning... it's an "amoeba" ;)
😃
Great tests, AI is getting better every day!!
Many AIs are great on generating unintentional humor.
yep, i had several lol moments from generations gone wrong
😂
I tired a few. This one turned out pretty good "A busy Tokyo market in the 1800s"
nice, thanks for sharing
Waiting for this stuff to get really good so we can generate fan series and movies for star wars and other destroyed franchises..
I really don't get why people are hyping 'AI' so much. Especially AI "art." Why do people parade other peoples' stolen work as their own and are even proud of it..? And yeah, let's cheer for the biggest wave of misinformation yet...
We are living the building of the primal stage of matrix
Thanks for the video, I appreciate it so much
No worries!
Server is already having a stroke
for sure!
luckily it wasn't bad when I used it today, took a few minutes and I got a creepy realistic video of a gray alien skulking around in the dark it was super eerie I loved it
@@wesley6442 same here. everything could generate within 2 min now. glad they fixed the server issue
I'll care as soon as I can run it local and we can train it. Still impressive though.
hopefully in a few months. you'll need one hell of a gpu though
@@theAIsearch I'm ready to buy a 5090. As long as it's not 3k€ lol
Really impressive, i am amazed at the video of the mafia guys fighting, the end with them stoping and shouting at each other is so cool kk, also, i searched in douyin (chinese tiktok) for kling ai videos but there is nothing new there, most is people talking of the videos that they released or comparing with sora and stuff, so i guess even with a chinese number you cant test yet kk.
thanks for sharing. i don't have douyin, but many users are sharing their kling generations on twitter. i wonder how they got access so fast
@@theAIsearch interesting to know that, also, douyin is kinda of a spywere kk so u dont lose much not having it, also, you need a chinese number to make a acount there too kk.
I'll just wait a few years. Not ready yet.
this makes zero sense. You are taking prompts from other tools which are certainly cherry picked for THAT tool, which is unreleased so we cant verify. Of course luma will produce worse. In my experiments, give Luma a source image and a dynamic camera and its in its element.
the dragon just snatch his balls 😂 less like video generator more like comedy
Love to see a comparison at 6 month and then a year from now.
Ai is never gonna replace humanity since it's just forever iterating its intrinsical inputs set by humans, and lacks external forces without human intervention. I don't know if it will ever be capable of producing truly novel ideas like humans can.. But maybe it's ideas are novel in their own dimension, akin to us being never able to truly comprehend the novelty outside our own dimension.
I don't think AI can never have a soul. It can only imitate, no matter how well it does so, we should never be fooled by it.
But in the end of the day, how do we know for sure what defines conciousness and soul? Why it should matter if brains are made of flesh and chemicals instead of metal and electricity, if they both have energy impulses that work in a similar way? We can probably never know for certain, but i think we should rely on our instincts.
My instinct: *BRO WHAT AM I SUPPOSED TO DO??*
Guys, Sakura only lasts about 2 weeks during spring 😅
for standard 150 prompts 30$ for pro 430 prompts 100$.
I don't know why they don't just use a conglomeration of programs to turn the scene into a 3D representation, then make the video similar to how games do!
You'd only have to create a few minutes worth of 3D space, not a whole game!
Bro I am extremely scared for the future. These stuffs are supposed to be impossible.
I would like to share some of my lumalabs videos with you, but I can't figure out how
the reason why ai does it is because it categorizing algo needs to be improved...it needs to know what a hand is & what makes up a hand and what those shapes are and etc. it's really just how they ai categorizes & processes data that can improve the generation.
THIS IS SO SICK
😃
I hope everyone just levitates away with infinite power, instead of being stuck in the confines of a society which can have that much power in hands
Any predictions on when we will see our first full feature movie that is in the theaters created written and fully created by AI
next year
No idea why but I can't really get it to generate anything really well. It often doesn't get even the basics right. I suppose there's a very certain style of how to prompt for the videos or something. Also, what do you mean that it generates two generations per prompt?... How do I check the other one?
Try with more detailed prompts - I heard this works better. It was 2 generations when I made the video, now they changed it to 1
it doesn’t understand physics or light, it just statistically guesses the configuration of pixels to getwhat you want. but it sure seems like it though. The real trick of AI happens in our mind.
A panda playing guitar is REALISTIC? LMFAO ;-P
Thanx for this sharing. Amazing tools. Does CHATGPT OPENAI can generate videos from input images ? Peace & love
you're welcome! chatgpt can't do that
@@theAIsearch Thanx for your feedback
But does I have character consistency
the first two props I tried wouldn't generate for moderation rules I would guess, already useless as an art tool. Commodification killed this thing already.
oh, what was in it?
@@theAIsearch I did the same thing I did when I tried out Haiper, was a grabbed a couple promps from nightcafe that I used to use alot. One was of donald trump the other contained the word butt. it told me it wouldn't generate them quick like for moderation, but it had no problem with bees in the trees. If they expect people to pay for this they need to a better job.
How do you set it to generate two at a time? Whenever I try it, it only creates a single video.
Me too... I came to the comments to see if I was doing something wrong. BTW, it's funny to use and if we realize that two years ago, all of this would've been sci-fi, it's impressive. Still, it's pretty much useless right now, it certainly needs training and fine tuning and that's probably why we're allowed to play with it for free.
@@dulume Yeah I'm excited to see how it improves over time.
it looks like they changed it now. i could generate 2 yesterday morning, only 1 now
Do you happen to know if paying the subscription speeds up generation? no because it's so slow and it's really impractical to use it..😅
i dont think it does
Kling you can use now without a Chinese number
Look closely at the “hands” holding the gun in the thumbnail. 🤣
?
@@theAIsearch one arm but two hands.
Make video on all video ganraor fight
As Chinese I found those two lunar New year clips extremely creepy and very uncomfortable to watch
lol
25:43 Why are we tokyo drifting?
First generation, 30 minutes waiting so far, lol... Still waiting 😆
Interesting, I'll try it. Thanks
enjoy!
The AI that learned from us seems to think we’re suicidal drivers. Should we be concerned? 🤔
well, some humans are pretty awful at driving. perhaps this contaminated the training data lol
Yeah just like the VHS
just imagine 10 years from now
At the speed AI is developing, it will be unrecognizeable from what we have now.
Still cant replace anime. Might help tho .
its not great right now, but img to vid might be useful for anime once they improve it
it only generates one video each time, not two. why?
they changed it to 1 now
I've managed to get a few generations out, but now the Queue is so long I've been waiting six hour and my next image still haven't popped up, I've did get an image to Video of an anime girl spinning around and when all over the place, Been dying to retry the image with her dong something with less movement.😬
yikes. try with a new account
Is it confirmed the Kling noodle guy is AI generated?
3 hours to generate a video? - It would be quicker to throw on a Panda suit and travel to the lakeside with your camera. Then amputate a finger to show people what good AI it is 😁
lol
It wants my credit card number first 🤑
you don't have 30 free generations???
I hope there will be a good video generator for adults
😃😃😃
How do you get two variations per generation? I only get one.
it looks like they changed it now. i could generate 2 yesterday morning, only 1 now
maybe you should make a discord server :)
Trying it now with my own picture. For whatever reason, I gave it a long promp that's ATLA fandom based.
....And it failed. Let's try again. (I wanna see if it even uses my face/body at all. Copilot hates faces because privacy. Screw privacy. We don't have any anyway. Everyone put their real names on FB, I mean, gawd.--I'm tired and not thinking clearly.)
It finished. It plays like a meme from a fever dream or one of those other AI videos that talks, but without sound. It pans down picture and there's ice under my feet, then it cuts to two women, with "me" on the right, talking, then fainting. Looked nothing like me except the skin tone and red lipstick, but she was wearing a blue dress like I requested, and a silver necklace.
Maybe the prompt wasn't specific enough, I don't know. Or it was too specific. I expected movie quality, instead I got a drunk meme gif that keeps looking and doesn't look like me. LOL? My Prompt:
the person in the picture being a water bender from Avatar the last Air Bender, wearing a long blue dress with silver embellishments and a necklace with a silver moon pendant on it. She has on silver fingerless gloves and knee-high grey lace up boots. She's glowing with power, standing on snow and waving water around in mid air, directing it in front of her.
---
Maybe it was just waayyy too detailed and long. But I write, I'm just word-y. Yep, that's probably it.
Hmm I'm not sure how complex your image is, but it's quite hit or miss. Try multiple times, or with different images
@25:54 whats going on with the car?😂
drift
Thanks for the hint🥳👊
No problem 👍