This AI video generator breaks Hollywood
ฝัง
- เผยแพร่เมื่อ 16 มิ.ย. 2024
- Runway Gen3 is here! Make realistic & consistent cinematic AI videos.
#runwayml #aivideo #sora #lumalabs #ainews #ai #agi #singularity
Thanks to our sponsor Wondershare Virbo.
Try Virbo for free: bit.ly/3xckK14
Download Virbo mobile app: app.adjust.com/1dd5k2bc_1dvobpnn
Subscribe @WondershareVirbo
runwayml.com/blog/introducing...
Review of other AI video generators:
Kling: • New AI Video Generator...
Lumalabs: • The BEST AI Video Gene...
Vidu: • New AI Video Generator...
Newsletter: aisearch.substack.com/
Find AI tools & jobs: ai-search.io/
Donate: ko-fi.com/aisearch
Here's my equipment, in case you're wondering:
GPU: RTX 4080 amzn.to/3OCOJ8e
Mouse/Keyboard: ALOGIC Echelon bit.ly/alogic-echelon
Mic: Shure SM7B amzn.to/3DErjt1
Audio interface: Scarlett Solo amzn.to/3qELMeu
CPU: i9 11900K amzn.to/3KmYs0b - วิทยาศาสตร์และเทคโนโลยี
Thanks to our sponsor Wondershare Virbo.
Try Virbo for free: bit.ly/3xckK14
Download Virbo mobile app: app.adjust.com/1dd5k2bc_1dvobpnn
Subscribe @WondershareVirbo
Review of other AI video generators:
Kling: th-cam.com/video/BTfLq-XkO0w/w-d-xo.html
Lumalabs: th-cam.com/video/jYAOPmMEyew/w-d-xo.html
Vidu: th-cam.com/video/FhCIciDRpj4/w-d-xo.html
We are a secret society that uses the second-best algorithm available, which faithfully and accurately replicates reality. No one in the outside world is aware of our existence since we maintain it under wraps. As of right now, all artificial intelligence in the world is terrible-very PRIMITIVE. WE'LL RELEASE OURS SOON.
LumaAi=Limited to No commercial use of their videos aka LumaAi owns your video.
RUnway=ZERO restrictions, you own the videos you generate
Runway wins hands down.
Except it is not released yet.
@@heww3960 Not Gen-3, but Gen-2 and Gen-1 are available and has no restrictions. Gen-3 is supposed to be out in just a few more days.
Luma offers subscription with commercial use
@@cathycronje4848 Exactly, subscribe for life or lose your commercial license, while Runway is free use for life. Runways value over a lifetime cannot be beaten.
Runway is going to be more censored
im so happy and excited at how FAST these new video models are being annouced and near releasing i was worried that sora level might take MONTHS or even a year but now with new video model stuff were just pretty closer to that reality faster!
it's crazy how so many video generators have come out in recent weeks
Competition is a good thing; it prevents big companies from releasing new features in a piecemeal manner.
It also makes big tech companies look foolish for teasing new models and then not releasing them. In a free market other companies will always fill the gap.
Yea. Let's put a pause on the announcements of "behind the curtain" tools. If it's not available it's not real.
If it's not real then what have they used to create the sample clips
@@robertanderson2424I think what they mean is, we should take its quality and practicality with a grain of salt
@@nickchua5772 oh aye I agree. Even the other dreaming one that just came out is not as good as the samples once you test it
Hallelujah! Everybody repeat after him and repost this under the comments. I'm so tired of this!
its not hard to build such model but the resource consumption is huge to accommodate users. so it is very challenging to provide such application on a monthly subscription
I've always hated Hollywood's productivity for being slow as hell. Maybe AI is finally here to fix this problem. It's expected to hasten the process of making movies which helps directors finish their masterpieces very quickly. Isn't it so cool to get a new sequel to our favorite movies every 6 months instead of every 3-4 years?
No
Soon we'll be sifting through tons of mediocre and low quality releases. We'll be drowned even more so than already. OTOH I think its a good thing in the hands of artists that otherwise could not afford this level of productions.
Faster faster faster!
What those AI models could be used immensely well are dreams in movies.
Imagine Inception using some of these dream-like features of AI, the likes that are near impossible to do by hand.
Yeah, I had a similar thought. They could just make movies featuring people dreaming, then use that technology.
so far Runway almost might be on the same level as Sora from what we are seeing so far.
Now comes the video length wars
how exciting!
2:52 fascinating :)
You can make amazing things with all this...
Everything depends only on imagination/ creativity (and critical and systemic thinking, but few people can do that).
market is heating up love to see it
I say about the sad truth. In Hungary
- Sora isn't available because we aren't on red list
- Kling isn't available because we aren't Chinese (VPN isn't option because you must have a Chinese mobile number)
- Google Veo isn't available because we aren't on the list (VPN isn't option because you must certify you live in a right country)
- Runway Gen-3 isn't available (I don't know why but I think it's not released yet)
Only we can use of the more modern AI text-to-video generators are Luma, LTX, Morph and Pika. But none of them can make longer than 10 seconds, mainly in good quality.
thanks for sharing! hope we get an open source model soon so that anyone can use it
The most difficult part of AI rendering video will be the last 1% of moving to true realism. And, it might well never happen.
Never is a long time. It will happen, even if it takes 200 years.
The level of realism is getting out of hand. I feel like where's getting reality 2.0 in my lifetime
At first I was fascinated by the train cabin controls, then I looked at the numbers and laughed. We'll get there.
Thanks for the video.
You are welcome!
What you think the next great leap after video improves, controlling the outcome improves and audio etc.
For this to be truly usable in rhe modern world it needs to be able to take video/world it creates and create that in a 3D modelles setting with textures etc matching exactly what's in the video
That way you could EASILY create a feature length animation/LED Volume scenery.
That would be game changing but would be destructive to jobs
Next phase will be trying to generate matching audio for videos. Imagine it trying to match piano audio to that scene... tough
BRAZIL MENTIONED! 🇧🇷🇧🇷
When is it going to be released? Ya teased us!
soon...
ai movies before gta 6 is crazy
Al is learning and a new art form is born.
exciting times!
There is no art here.
"learning" lol
@@beanieteamie7435womp womp
Makes you wonder if one day we will have real-time AI generation which will replace 3d modelling. Everything will just be a generated real-time video.
This was really good, i think it is one of the worst in mantaining deatils, buuut, it is the best in looking amazing, it is simply insane how cinematograthic it is while being so inconsistent.
Yes, I like the cinematic look too!
Bro is my daily dose of ai
Thanks!
Until we actually have access to sora, we should stop using it as a benchline for quality because it could be totally fake for all we know.
Some have gotten access to sora. Maybe he knows one of them
@@Grassland-ix7mu
The classic "I know a guy".
It's a new era. Things get more exciting and also scary now.
what a time to be alive!
I have yet to see an AI-generated piano keyboard that looks correct and realistic. Most of us know the black keys occur in a pattern of 2, then 3, etc. You can see at about t=16:00 that is not the case here. Musical instruments have been difficult for AI generators. I am finally seeing violins with the bow straight and resting on the string--and the proper number of strings.
Exciting times
yes!
People aren't appreciating how ridiculous this AI is enough. It is actually way better than Sora in the creativity aspect. The visual effects it makes are astonishingly good.
If you think of AI-generated media, it's really just the ultimate level of video compression. You could save so much bandwidth by just streaming 260p video and your local ai-powered tv set can interpolate frames to 1000hz, and upscale the resolution beyond 10k.
The bandwidth could be used for so much more powerful applications like networked simulations and data mining (not always a bad thing). Most of the traffic online is streaming video and audio.
Or you could just tell your tv to generate video, completely independent of an internet connection, and just watch an endless dream...
very interesting idea
I'm assuming the stepping in motion on many clips is actually your mismatched frame rate. It would help to change the clip frame rate to match your edit timeline. My guess is the clips are 24fps but your edit timeline is 25fps as there is the classic frame repeat every second. Do the AI generators give you frame rate options for the rendered output?
🇧🇷🇧🇷🇧🇷🇧🇷👏🏻 Wow, Runway V3 is amazing! It's not yet better than Sora, but it's pretty close. In my opinion, from what I have seen.
agreed!
It looks like runway generates 10 second against the 5 seconds in Luma
I can see world model in the future from here.
amazing 😊
😃
This month is the month of ai video and future months
yeah, so many video generators announced!
They should let the public know when this stuff is ready to be released. Sora had everyone screaming few months ago with that commercial. Today, Sora still doesn't exist.
"Generate ANYTHING that says ... on it and/or have straight lines." Why all of them fail is a mystery to me.
Now I know how The Matrix will generate all the imagery to feed to our brains.
At 14:48 the girl goes from not having a side-nose piercing, to have it, which then disappears, while a nose ring appears instead.
Also, at the start the girl looks Korean-Asian, to then look Indian.
In 5 years anyone could make a Hollywood movie.
or next year
Well, AI is already capable to producing scripts. It's only a matter of time before it actually creates the scenes, actions, and dialogues for said script!
If you are the type of person who could come up with an idea that strikes the balance between being original enough to be interesting but generic and relatable enough not to make people uncomfortable and understood how to make convincing dialog and know how to plan a story arch for the length of a feature movie, can create tension and release in multiple phases until a final satisfying climax and resolution of the hero's journey while making some sort of statement or point and the only thing holding you back from making a movie was the fact that you could not create a corresponding striking visual image; then yes, you will be able to make a good looking movie soon. But they already have a name for those people. They are called writers. I don't get why nobody thinks about what makes something good. Is your favorite movie the one that looks the best? If everybody can do something easily why will anybody look at that thing or care about it? The only thing that is worth your attention (besides your friends and loved ones) is something that NOT EVERYBODY CAN DO, not what is common, and any person can make. Money as the main value mucks some of this up because people think that something good should cost a lot of money and that makes it valuable. But it is really scarcity that makes something valuable. Why would a diamond ring make a good wedding ring if your gravel was diamonds? Great movies mostly are made by a whole team of creative people doing what they have focused on getting good at for their whole life. Previously that took a lot of money to facilitate that happening. And most people do not have that kind of money. That is where the scarcity comes in and makes that value. If it takes little to no money to make a movie it won't automatically be a bad movie, but if it only has the input of one person it will likely not be as good of a movie as one that has hundreds of creative professionals specializing in one aspect of the creative process. The combination of that creativity makes a magic that can change your life, your perspectives, and make you cry and laugh at the same time. Prompt that. If your tools, in a matter of moments, can do everything that humans used to do, then the situation has changed to where the tool is not a what, but a who. If YOU are not adding value to a creation, in some fashion, I suggest that YOU are actually the tool who is being used. I've trained my newer and faster replacement at jobs before so I can see a pattern here. I don't mean you personally @69mustang, I mean we as humans and that is just how I feel about the situation we find ourselves in currently and food for thought!
@rundajulesproductions7735 I'm with you 100% on your statement. 👍
yea right just because you can create movie easily doesnt mean its gonna be good or that anyone will wanna watch it
"hyper-fast" seems to mean a different thing to the AI than it does to me
Notice that all the choices in the prompts avoid any locations or props where signage and typography would be visible. We accept graffiti as being illegible.
good point. hands & fingers were also avoided
16:22 Lol agreed!
Impressing, but it is not released yet. Would want to see a eating video.
But... can it generate good porn? ;-p
Just ask ai to render hands. It never broke hollywood
5:27 That's not how lens flare works. The warped motion of it is all wrong. The discs should like up with the light source too, not move around independently.
when can expect to see video generation with sound?
not sure. a workaround for now is to describe the video in udio/suno/elevenlabs to generate some sound fx
I think if the speed of development remains consistent it could be even just 1 year until we see sound and massive improvement in the video
Yeah, I'll need sound to generate AI porn... porn is always an early adopter in tech... ;-P
@@igorschmidlapp6987 true haha. I feel like they set a lot of restrictions on the AI at the moment though, most don't want the bad name just yet. Once it's good it will get monetized more lucratively. Then you'll get the P*rn
@@igorschmidlapp6987 we'll need to wait for opensource for this. runway will be censored 🥲
Until they give us access to the tool , I’m highly convinced that they trick us into believing in some of their “ realistic “ generated videos. It could be real life footage who knows 🤷♂️
Just your imagination
The exit is your fixation
Door after door you find that is just a *digital hallucination*
You try to find a way out
but you're stuck here day After day insiiiide a
*DI-GI-TAL HAL-LU-CI-NA-TION*
luma labs dropped down to 10 films per month now since so many people went to it.. so you have to wait a month now :(
it is not slightly but STRONGLY BETTER THEN LUMA AI
BRAZIL MENTIONED!!🇧🇷🇧🇷
YEAAAAAAAH MANNN!! ANHOTER ONE !!! OOOOOOOOOOOOO!!!! YEAAAAAAHHH!!!!👏👏👏👏👏🔥🔥🔥🔥🔥🔥🔥🔥💯🗣🗣🗣‼️‼️‼️‼️‼️‼️🔥🔥🔥🔥
🔥🔥🔥
@@theAIsearch I luv uuuuuuuu
its the best so far, just 4s the problem
i hope it's 10s like the examples, and not 4s like gen2
yes the girl looked emma watson
Is your voice in the videos also Made with AI?
no
@@theAIsearch😂😭
I’m over it already
piercing appearing on the nose at 4:12
👍
I'd be extremely impressed if it is free.
i doubt it. runway has been quite expensive in the past
Do you know what midjourney is up to? A few months ago I heard a leak that they are working on some sort of 3D model or Image to video that’s supposed to be “insane” but ever since I heard absolutely nothing. Do you or anyone else what they are up to?
not sure. there have been rumors here and there, but nothing substantial
the CTO is not satisfied of their current video performances, but 3d may come early
@@sherpya thanks. Where you got this info from?
@@elonmusken9513 a guy on TH-cam summarizes mj office hours, the channel is called future tech pilot
@@elonmusken9513 office house, a guy on TH-cam summarizes each session, I can't write the channel because youtube keeps censoring me
hey dear
there's a youtube channel named MIGHTY MONK , this channel probably uses RVC or TTS for generating voiceover, can you make a tut on it, you have made in the past but now can you make again like updated one cause ai stuff has been changed n improve so can u?
i personally wanna generate voice like that channel do create.
thank you , love from india
Google is completely overwhelmed, such an under performer, only talking not releasing. What a loser it has become
Break what. man? Hollywood? That shit breaks Hollywood? 😂😂😂😂
can we Access it?
14:48
Runway Gen-3 Is little bit worse than Sora but faster in movement and has big potential
Its just a sad thing that rise of AI tech has to happen when socio-politics on the west is at its bottom state, just telling you what kind of dystopia we're heading.
Consistency will come overtime as they get more data to (not generate but to reuse)
So in order for a street to be perfect, they have to have perfect data associated with each object or a footage of that street matching exactly what you typed in the prompt.
I am starting to see how this works and while its impressive and all, its just a REUSE and not GENERATE or CREATE something new.
but nobody watches movies anymore.
Are you confusing the words "macro" and "micro"? Macro = large. Micro = small.
macro photography is close-up photography of small subjects
@@theAIsearch Interesting, I guess I learned something today, thanks!
But will it be censored? If it's censored then I'm not using it.
yes
@@theAIsearch Bah, fair enough.
so you have no idea how far is kling to this lol
"Simple scene" - lol try to create the whole pipeline for this kind of magic like video footage generation yourself, lets then talk about "simple" :D
Salut. Il y a énormément de défauts dans les vidéos montrées... les rails du trains, les termites qui sont dans une pelouse 15 fois trop grande pour les maisons environnantes... heureusement que les détails restent flous. Pour le graffiti, en arrivant il n’est pas sur le mur pourtant on le voit se matérialiser avec une vue arrière. Et la nana tatouée, son soutien-gorge n’a qu’une bretelle qui deviennent deux, etc, etc...
Je trouve ça joli mais il y a tellement de défauts que ça en devient ridicule si on veut le comparer avec un travail humain. Je sais que la différence sera minime dans quelques temps. Alors que toutes les deux semaines, un nouveau venu dans l’IA prend une place dans ce domaine, les avancées sont exponentielle et presque tout les jours une annonce nous montre une nouvelle manière de l’utiliser ou de nouvelles possibilités...
On avait pas tout ça quand je suis né, pourtant à 51 ans , je n’ai pas l’impression d’etre obsolète.
Au plaisir.
Not open source, not interesting.
15:12 - Wow, if someone showed this girl's video to me I would have thought for sure it's simply a real video of a real woman and not AI gen'
yeah some of these are hard to tell it's AI if we don't look closely
@@theAIsearch Yes, so crazy, since this is still just the beginning of it, like give it a few years and they would most likely perfect it all out so then even the most tricky prompts would look 99.9% real as well without a single error.
Anytime I see "z killer" I know it's overhyped bullshit, can't tell you how many times I've seen something from Google be called a killer yet it's some the worst ai hands down, better off running things locally, or at best it's a tie.
Ive been using gemini1.5 for a couple of days and its hands down been the best at the task of translating a c++ game dev book into python. Agree their demo was sooooo slimy and misleading though. Just like the majority of advertisers and politicians, promising all kinds of things to get those 💵
sentdex gta5
no, it doesnt break hollywood..
How terrible that you can make only 10 seconds video... In Hollywood, a film's length is between 70 and 130 minutes...
Not trying to be "that guy" but When describing something very close up it's Pronounced microscopic not macroscopic
macro = close-up shots of small subjects. see macro photography
@@theAIsearch oh touché!
The moment I lost feeling for AI when I saw that any AI that we use to generate something, AI always generate based on the existing pattern. It doesn't create anything new. It often faces hallucinations. So I'm very disappointment like how people are spreading false rumor like AI took the job from them
" It doesn't create anything new." - That statement is so untrue, I can not believe you even wrote it!
It? Which ones have you tried? When did you try them?
@@6AxisSage There are literally hundreds of various quantized AI Models to choice from. Each one has unique strengths and weaknesses.
@@jdsguam yeah, thats why im asking. Ive talked with people who reference free chtgpt when it was gpt3.5 as the one interaction with expectations of 4. The argued with a reflection of themselves and that and doomer pr_ on is their only understanding. If you treat a llm as a next word predicting machine that keeps on going. Then look at the benchmarks, how accurate is that model? Test the model, learn its emergent behaviors. Did I manage to leverage something useful or gain insights? Then I try newer models the day they come out and try those.
Is it wierd i prefer the fight or the model which is capable but refuses over an agreeable local llm that does anything I can imagine? Like the reflection is too pure?
@@6AxisSage I've download many models using the LM Studio UI. I've jailbroke a few; but, I've not argued any yet. lol
First
😃😃😃
Am I the only one not finding this useful ?
I am with you.
maybe for some part of a music clip
Think abstract here. Currently this is more a tech demo, footage from a prompt. Imagine a day where there can create a 5m, 30m 1hr video. You are seeing a step of progress
I'm finding many uses. I'm working on music/lyric videos and short films
Not me, very useful stuff!
New interface? Or same lame cumbersome interface as always?
Anime artists are probably irritated if their work has been used to train ai without reimbursement.
Except it’s impossible to reimburse every artist. Plus, diffusion models and even transformers, don’t just copy & paste, and instead they “learn like humans do” in a sense. That’s why Japan has already declared that copyright laws don’t apply to AI training, quite some time ago.
I’m confused, why just mention anime artists?
Said the Anime artist that had his work used, copied, inspired and manipulated by other aspiring Anime artists.
And where did anime artists get their inspiration from? No one owns art
@@mariomills If you had to spend time, money, mental energy and physical energy to learn to make professional art and then train your AI strictly on your art, would you do it? I'll bet the answer is no. I'll also bet that even if you did, you would never give your AI to six million users for free, who would then drastically reduce your chances of making money with the skills that cost you so much money to obtain. Why then should anyone do that for you?
Millions of useless Videos incomming.
Yes its a cool Instrumemt. Like the Trompet. Give everbody a Trompet 😅
This model is mainly for public use I think, so it's not that powerful as Sora.
It all looks fake 😂
for now
Luma sucks. Every video it generated for me is terrible.
"Of course Hollywood will going to be really pissed by this" By Hollywood you mean the top executives? Oh they will love this because it would mean more for less money. The people who are going to be pissed by this are the hard working effect artist and cinematographers etc, whose work was used yet again without their consent and any kind of compensation to make a tool that will be competing with them. The poor excuse that this is ok because of democratization of certain fields is such a bullshit.
Kling ai can also be used, but it can only be used in China at present 😂
Yea. Let's put a pause on the announcements of "behind the curtain" tools. If it's not available it's not real.
First
😃😃😃
@@theAIsearchIdk why but this emoji terrifies me lol
@@tentzz you probably saw the movie Smile. It is the same horrific smile. 😆