Gen-3 Alpha Text to Video is now available to everyone. A new frontier for high-fidelity, fast and controllable video generation. Try it now at runwayml.com
@@Oldyellowbrick A videographer wants to portrait something real that you can relate to. As more AI you watch as more strange it is that nothing is real, no history, no culture, just things that barely reminds of something.
wowh ... just wowh ... I am dealing with A.I. since 2016, and we tweak around with A.I. for our business now since months, already very happy with to push our business model further front and far ahead the competitors. - If the answer to the question : Is every sequence in this promo trailer generated A.I. is "yes", then its mind boggling ??? Luckily we have experienced the "old world" = 100% reality, grown up with b/w TV in the 60s. I am sorry for the new generation growing up with this "fake world". For businesses, yes, its fantastic. For the human psyche I have my doubts we - as human race - move into the right direction. The emotions we can create with A.I. can tear us apart ... the risks for mental damages are huge. Doing Internet business for more than 30 years, have we overstepped a line ??? Its a serious question about mental health care we have to address.
This just in, generating video costs tons of compute (=money) and probably most the reason sora is still private is because of the insane compute it also costs.
SORA is staying private until after the US Presidential election. they don't want to introduce high resolution deep fakes into this already crazy US election
Dont worry ... the investments into data centers, see latest A.I. data center for MIcrosoft at around 3 billion, will be re-financed eailys by the huge processor speed. Imagine, that one GPU unit of Nvridia's new A.I. specific Blackwell can handle 1.8 million calculations per second, and its just one unit. A data center is built with thousands of such units, and can operate billions if not trillions of operations per seconds. And service millions of customers. Alone on TH-cam more than 2.49 billion people in march 2024. Just imagine that huge number. Its nearby 1/3 of human race watching vids. So there is no need to worry about re-payment. The challenge by sure is the pre-investments and to get this financed. But there are trillions out there of capital which seek lucrative investments and by sure A.I. is one option, investors, hedge funds are happy to put their capital into it. It will pay out easily.
What you have just witnessed isn't merely the demonstration of AI's fidelity when it comes to creating realistic imagery, it's much more than that. It is the distant prelude to The great filter, the explanation to Fermi's paradox, humanity will never reach stars because it's gonna be completely replaced by AI and virtual realm in which it will reside. A selfsustaining entity with little or no interest in the external world. This video is equally facilitating and frightening.
1920x1080xRxGxB start from black all pixels in active. All the way up to white all pixels active and infinity ♾️ in between. Well done Runway there is no ways this was easy.
I just created a zombie film using Gen-3 and once I got used to prompting like a film director, the results were great. Still far from perfect and way too expensive, but it's only getting better (and one day cheaper).
You dont know how many hundreds of employees they needed to get that promo thing done. Project manager: Guys, today we are 270 people in the work shift. I expect from you per minute at least 5 new prompt text ideas to get a 3 second sequence done. I give you in the next 4 hours till midday break time to generate 81,000 prompt variants, then our A.I. will analyse all ideas and pick out the top 10. The best 3 prompts will get an extra bonus today of 500 bucks. Ready to go ? Time runs now! (presses the button)
Great that this is released! Burned through my credits pretty quickly but definitely some beautiful moments here and there. Looking forward to price lowering quite a bit before we can really get into full experimentation mode, but this still feels like an important step forwards. Thanks for all your work.
promt: Three gray pyramids with futuristic details floating in a snow forest with blue fire below recording made by the arri Alexa lf camera and the Mercury atlas lens with dramatic and warm colors slightly blurred scene and camera shaking a little
Now share the prompts used to create these. Chances are these were iterations upon iterations. For one person to get that result will likely cost them all of their credits because the initial prompt, while detailed, may not perfectly match what Runway's modeling is looking for as an input. Half of the problem with using these tools is learning how each model wants to be prompted in order to create the desired output.
as a producer and artist i've been using gen3 for real world artistic purposes. My most recent single's music video was created with gen3 and for the first time ever it's actually worth it. I spent about $30 in credits. I'll be making more music videos with it if anyone wants to follow along
@@christianb8900 With the $15 monthly plan on Runway, you get approximately 1500 credits per month. Since generating 1 second of Gen-3 content costs 10 credits, you can generate around 150 seconds (2.5 minutes) of content per month with these credits. If you run out of these credits, you can purchase additional credits at the rate of $10 for 1000 credits. This additional purchase would allow you to generate another 100 seconds (1 minute and 40 seconds) of Gen-3 content. So, combining them for $25 150 seconds (2.5 minutes) from the initial monthly credits 100 seconds (1 minute and 40 seconds) from the additional purchase This totals approximately 4 minutes and 10 seconds of Gen-3 content. So imagine the content you generate isn't up to your ideals for half your generations and you want to make an actual full episode of something 20-60 mins, going to cost hundreds of dollars/maybe even thousands. Not sure if the unlimited generations for $95 a month will be worth it, if gen3 takes hours to generate something (they mentioned a more casual slower generation mode for unlimited) and when there's millions of users, could take weeks before you get 10 minutes of useable content. $15 is not really an accurate representation of how much it costs if you are serious.
@@christianb8900 But you only get 62 seconds of video generation a month. So this particular video would take 3 months ($45) to create assuming you never reroll or edit or trim any generated video and don't cherry pick anything and you must use every single generation or it will cost more and take even longer.
I'm so glad we gave the keys to the Library of Alexandria to some tech bros who giggle like little Dutch schoolchildren when they see some keys dangling in front of them. Now we get to enjoy visuals that aren't bogged down by the pesky limitations of weight, framerate, continuity, texture, perspective, momentum, or taste. And it only uses enough energy to burn down 4 houses for each video you generate (not counting all the times you start over because the armor on the super busty, caked-up wood elf isn't showing the clearly defined nipples that you clearly defined in the prompt.) The future is here 🤩
Taco, you are born to become a comedian. Come on. Start your own podcast. I want hear more ... :-) We need critical voices (see my upper comment). Yes, its insanity ... but who cares. Insanity drives us humans since thousands of years, isn't ?
What books do you recommend about the mind? I like how you always leave on a positive note, letting me know they are creators in how they control their own thoughts and what they want out of life.
@@that-one-music-guy Yes you’re right. I have had the luxury to witness beta through a friend and it did not impress me much. It is amazing piece of tech but the videos you see have been perfected over many trial and errors. So if our do get it make sure you get unlimited
@@iKINDREADOh ok I see, thanks for the clarification, I mean it makes sense, there obviously going to cherry pick for there advertisement, and I'm sure they've spent countless hours in post,
The facial generation is unbelievable! Amazing, impressive and breathtaking - yet simultaneously rather disturbing. Looks like we've finally bridged the Uncanny Valley and now I'm not sure I want to cross it. The repercussions for society are going to be profound.
without transparency around training data all I think about when i see new AI video work is the thousands of videographers, travel vloggers, drone operators, archaeologists, scuba divers, uber drivers, soccer moms, etc. all the video equipment they bought, the time getting the angle right. the color grading they paid their friend to do. the little lighting tools they bought on amazon. the weekend scanning family photos with grandpa. its wearing on me
I'm not saying that AI is theft, but the impact on society could be compared to the same as if digital theft was allowed from one day to another. edit: but funny enough big companies pushing for AI to cut costs will eventually also suffer, then they will try to take monopoly on AI.
@@aymanbenbaha yes, of course those responsible for creating the work in the dataset can continue making videos, they will be unknowingly be used for new datasets GEN-6 or whatever. this doesn't remove my concerns about data secrecy and shameless capitalism and the normalization of exploiting intimate personal moments from people without their consent. I'm not anti-AI and I use it for years and train models of my own from time to time but just not so excited about perfect video generations that I can overlook how its made and sold
@@Champignon1000 yeah it can be theft for sure. in certain contexts it's mathematically calculated theft to avoid copyright and leapfrog the process at the direct expense of those who didn't. sounds theivy right lol
@@incription ikr, I was thinking the same exact thing, I need at least to generate 5-20 minutes of content to get a feel for it. Needs to at least give 10x the amount of generations...*sighs*
It's impressive, but you don't give any possibility to test it and know if we will get performance out of it. Gen-2 uses 5 credits per second of video generated. Gen-3 Alpha uses 10 credits per second of video generated. I have 10 credits in my account, how is it possible that I can do something viable in 1 second? How am I going to practice with one second? The basic plan is $15 a month, that's 62 seconds of generations per month. And that's assuming that it gives you what you ask for the first time, because if he doesn't give you what you want, you'll blow your credits just by learning. From my point of view, I am not going to pay a subscription if I can't even know how to take advantage of it. It is a pity. You should consider giving carte blanche for a while to people who create content using artificial intelligence, because there are many platforms that offer a good trial period, and a new one is discovered every day. You are great, but I can't test you, therefore I don't know what results you can give me. If I have a trial period and it works for me, I will probably subscribe, but not otherwise. Honestly, your application is the one that I have tried the least,and I have not visited you again. But you are great, and thanks for this video! all the best 👍💪
Now I am starting my own movie production house, I will produce movie for myself, I will watch myself and I will give Oscar to myself.
show, don't tell... no one cares
😂 Perfect
very good kkkk
Maybe you can even create an AI version of yourself to slap yourself.
hahah waah nikhil sir 🤣
What a time to be alive
Next time will be better
..unless you’re a videographer
2 more papers down the line this is gonna be crazy
@@Oldyellowbrick A videographer wants to portrait something real that you can relate to. As more AI you watch as more strange it is that nothing is real, no history, no culture, just things that barely reminds of something.
What a time to be unemployed
Some of those faces look 100% real. Insane how far runway has come in only 2 years
Doesn't render that quality. You will need upscaling for sure.
because it just steals
@@mbobby99 Sounds like someone knows nothing about how the tech works. Keep telling yourself that though, maybe it'll help you keep your job lol.
Yeah, from useless to useless in 2 years. What a progress 😅
@@iKINDREADthey literally said no modifications
crazy stuff, it's amazing to see how far Runway has come! congrats team.🎉
wowh ... just wowh ... I am dealing with A.I. since 2016, and we tweak around with A.I. for our business now since months, already very happy with to push our business model further front and far ahead the competitors. - If the answer to the question : Is every sequence in this promo trailer generated A.I. is "yes", then its mind boggling ??? Luckily we have experienced the "old world" = 100% reality, grown up with b/w TV in the 60s. I am sorry for the new generation growing up with this "fake world". For businesses, yes, its fantastic. For the human psyche I have my doubts we - as human race - move into the right direction. The emotions we can create with A.I. can tear us apart ... the risks for mental damages are huge. Doing Internet business for more than 30 years, have we overstepped a line ??? Its a serious question about mental health care we have to address.
RIP Sora. We hardly knew ye.
Is not free bro 🤣🤣 iam literally wating 1weeks
nah sora is better'
this is still very good tho
No. It's not even close.
Both are good... And both are NOT free.
Unbelievable. Hats off to the Runway team! 🔥
This trailer is something else...Fantastic. Congratulations on the release and congratulations on the trailer, it is really really good.
Tears in my eyes thank you team
If you create a vid of a control room, is there a way to make the camera not move? I tried entering tripod shot but the camera still panned.
Congrats! Looking forward to using this on some projects!
This is starting to get insanely good.
This just in, generating video costs tons of compute (=money) and probably most the reason sora is still private is because of the insane compute it also costs.
SORA is staying private until after the US Presidential election. they don't want to introduce high resolution deep fakes into this already crazy US election
it's not private you can get it with a subscription
Dont worry ... the investments into data centers, see latest A.I. data center for MIcrosoft at around 3 billion, will be re-financed eailys by the huge processor speed. Imagine, that one GPU unit of Nvridia's new A.I. specific Blackwell can handle 1.8 million calculations per second, and its just one unit. A data center is built with thousands of such units, and can operate billions if not trillions of operations per seconds. And service millions of customers. Alone on TH-cam more than 2.49 billion people in march 2024. Just imagine that huge number. Its nearby 1/3 of human race watching vids. So there is no need to worry about re-payment. The challenge by sure is the pre-investments and to get this financed. But there are trillions out there of capital which seek lucrative investments and by sure A.I. is one option, investors, hedge funds are happy to put their capital into it. It will pay out easily.
Is there a way to access the prompt necessary to generate this video?
Sales video project starting this month so the timing couldn't be perfect! This will definitely make me subscribe D
can we have the prompt for each footages?
Will it cost us a Kidney and make the lights dim in a entire state to make a hour long movie?
We need image to video with this model, so we can get exactly what we want
An article I just read said image-to-video and video-to-video would be implemented "in the coming days".
What are some legitimate use cases for this?
It doesnt work when i click on it, it sends me to gen 2
What's the music name?
I CAN'T WAIT TO USE THIS!!!
what is the
maximum output resolution ?
720p unfortunately
Can provide prompt for each segment. @Runway
Incredible results, thank you ❤
those eyes close shots.. were amazing
Hey guys, just a small tip, if you put the https before your link in the description, it will make it clickable. Might help with more link clicks
Congratulations!!
One question, can Gen3 Alpha make videos from a certain image?
And, can the same character be maintained in different scenes?
It won’t be long before this technology is used on the big screen!
What's the BGM name?
You guys are phenomenal ❤ thank you so much 😊
No way it's insane 🤩
Complex prompts doesn't work....
What you have just witnessed isn't merely the demonstration of AI's fidelity when it comes to creating realistic imagery, it's much more than that.
It is the distant prelude to The great filter, the explanation to Fermi's paradox, humanity will never reach stars because it's gonna be completely replaced by AI and virtual realm in which it will reside. A selfsustaining entity with little or no interest in the external world.
This video is equally facilitating and frightening.
lets be real. It looks like sht, moves like sht, and heavily limited
When it can generate 1080p footage and offer a RAW/LOG version of the files, we'll have something truly remarkable.
1920x1080xRxGxB start from black all pixels in active. All the way up to white all pixels active and infinity ♾️ in between.
Well done Runway there is no ways this was easy.
The amount of work involved here. What a video. Bravo 👏👌
I got shivers as a filmmaker. Goosebumps babe!
Ok, Will be make vídeo from Photo?
I just created a zombie film using Gen-3 and once I got used to prompting like a film director, the results were great. Still far from perfect and way too expensive, but it's only getting better (and one day cheaper).
This is incredible! I can’t wait to see all of the magnificent pieces of art that will come from this!
"art"
@@Petrovelly Cope
@@FreshPickedLemon Opinion
This is will revolution the pr0n industry
we need video to video for that , awsome wow
i love this
I’m trying it out and 99% of the prompts that I get don’t look nearly that good
You dont know how many hundreds of employees they needed to get that promo thing done. Project manager: Guys, today we are 270 people in the work shift. I expect from you per minute at least 5 new prompt text ideas to get a 3 second sequence done. I give you in the next 4 hours till midday break time to generate 81,000 prompt variants, then our A.I. will analyse all ideas and pick out the top 10. The best 3 prompts will get an extra bonus today of 500 bucks. Ready to go ? Time runs now! (presses the button)
that's whole point of doing it yourself no? if you have a braindead prompt of course ur going to get a braindead reply
Would be interesting to see the whole prompts in written Text Form
I am so happy, that I've subscribed, can't wait to try it.
Great that this is released! Burned through my credits pretty quickly but definitely some beautiful moments here and there. Looking forward to price lowering quite a bit before we can really get into full experimentation mode, but this still feels like an important step forwards. Thanks for all your work.
does it have image to video?
@@DarionDAnjou doesn't seem to, that was what I was most eager to try.
@@aiforcultureai video results always seem to come out better when to starts with a still image
- Вы испугались?
- Обосрался
Impressive details, I guess the Gen-4 would be hard to distinguish from real world.
1:07 made me chuckle
why?
omg we're at the beginning of real cyberpunk
GoPro has left the chat.
promt: Three gray pyramids with futuristic details floating in a snow forest with blue fire below recording made by the arri Alexa lf camera and the Mercury atlas lens with dramatic and warm colors slightly blurred scene and camera shaking a little
Insane improvements in ai video generator that it's scary. Btw, you guys shoulda used can you hear the music for background
Now share the prompts used to create these. Chances are these were iterations upon iterations. For one person to get that result will likely cost them all of their credits because the initial prompt, while detailed, may not perfectly match what Runway's modeling is looking for as an input. Half of the problem with using these tools is learning how each model wants to be prompted in order to create the desired output.
Now image to video! cannot believe this wasn’t shipped with that feature.
Impressive. Very nice.
Now let's see the hands.
yeah! you guys are awesome!
Okay, now I'm terrified.
Can i make anime with gen-3 ?
Share the prompts please, so we can learn quicker!
You can view some prompts on their website
很棒
as a producer and artist i've been using gen3 for real world artistic purposes. My most recent single's music video was created with gen3 and for the first time ever it's actually worth it. I spent about $30 in credits. I'll be making more music videos with it if anyone wants to follow along
Awesome
fantastic! can't wait to generate music videos for our up-coming AI songs🙂
Finally, I'm making hollywood grades with this one! 🎉
You need money
you are really not
u won't hahahahahahahah
@@naturesoundup u cant even afford it
@@kayes4495 don't judge a book by its cover, eblan
This ad is Ai saying “Hello, can you see me now?”
we’re witnessing history right here
My first thought wasn't "Wow, amazing" it was "Hmmm, I wonder how much a month this is going set me back?" ha ha
$15.
@christianb89 yes I have no money
@@christianb8900 With the $15 monthly plan on Runway, you get approximately 1500 credits per month. Since generating 1 second of Gen-3 content costs 10 credits, you can generate around 150 seconds (2.5 minutes) of content per month with these credits.
If you run out of these credits, you can purchase additional credits at the rate of $10 for 1000 credits. This additional purchase would allow you to generate another 100 seconds (1 minute and 40 seconds) of Gen-3 content.
So, combining them for $25
150 seconds (2.5 minutes) from the initial monthly credits
100 seconds (1 minute and 40 seconds) from the additional purchase
This totals approximately 4 minutes and 10 seconds of Gen-3 content.
So imagine the content you generate isn't up to your ideals for half your generations and you want to make an actual full episode of something 20-60 mins, going to cost hundreds of dollars/maybe even thousands. Not sure if the unlimited generations for $95 a month will be worth it, if gen3 takes hours to generate something (they mentioned a more casual slower generation mode for unlimited) and when there's millions of users, could take weeks before you get 10 minutes of useable content. $15 is not really an accurate representation of how much it costs if you are serious.
@@MR_AK558 have you tried work?
@@christianb8900 But you only get 62 seconds of video generation a month. So this particular video would take 3 months ($45) to create assuming you never reroll or edit or trim any generated video and don't cherry pick anything and you must use every single generation or it will cost more and take even longer.
It's not free though...
When does this cross an ethical issue.
Shortly after Sora’s reveal, Cristóbal Valenzuela was asked if Runway will reach Sora’s level. His response? “Better.” He wasn’t lying, it seems.
Android?
I'm so glad we gave the keys to the Library of Alexandria to some tech bros who giggle like little Dutch schoolchildren when they see some keys dangling in front of them. Now we get to enjoy visuals that aren't bogged down by the pesky limitations of weight, framerate, continuity, texture, perspective, momentum, or taste. And it only uses enough energy to burn down 4 houses for each video you generate (not counting all the times you start over because the armor on the super busty, caked-up wood elf isn't showing the clearly defined nipples that you clearly defined in the prompt.) The future is here 🤩
Taco, you are born to become a comedian. Come on. Start your own podcast. I want hear more ... :-) We need critical voices (see my upper comment). Yes, its insanity ... but who cares. Insanity drives us humans since thousands of years, isn't ?
I am a time traveler and I am allowed to say that THIS IS THE BEGINNING OF OUR DESTRUCTION
What books do you recommend about the mind? I like how you always leave on a positive note, letting me know they are creators in how they control their own thoughts and what they want out of life.
Absolutely amazing, thank you
Congratulations…now is the time to push harder than ever
RIP Sora
too expensive for private use
well at least when all the Animals are extinct we can still make doco's about them
wow!
I am amazed and incredibly happy.
Like with Luma AI, they only show the good parts until you actually test them out
damnnnn, the close up shot, real threat for other AI
Not really tried for a month. It is not as good as promoted
@@iKINDREAD I think this came out 6 hours ago so trying it for a month would be imposable unless you had the beta, I think you tried gen- 2
@@that-one-music-guy Yes you’re right. I have had the luxury to witness beta through a friend and it did not impress me much. It is amazing piece of tech but the videos you see have been perfected over many trial and errors. So if our do get it make sure you get unlimited
@@iKINDREADOh ok I see, thanks for the clarification, I mean it makes sense, there obviously going to cherry pick for there advertisement, and I'm sure they've spent countless hours in post,
RIP Luma
No bro luma is free
Luma is also good
Hmm, A Magnific like Video Upscaler will Change the Game!
Can you get the video to loop?
Uhhh is this new? Extend 4Sec Video clip? Thats awsome Thanks keep it up❤
this might just be better than Luma
I'm glad that this is available now, but is it strictly text to video and not yet support image to video?
Amazing I worry for the legacy media video generation.
Wow this is more realistic than Avatar 2.
I'd love to see the prompts that created these shots. Amazing work!
The facial generation is unbelievable! Amazing, impressive and breathtaking - yet simultaneously rather disturbing. Looks like we've finally bridged the Uncanny Valley and now I'm not sure I want to cross it. The repercussions for society are going to be profound.
without transparency around training data all I think about when i see new AI video work is the thousands of videographers, travel vloggers, drone operators, archaeologists, scuba divers, uber drivers, soccer moms, etc. all the video equipment they bought, the time getting the angle right. the color grading they paid their friend to do. the little lighting tools they bought on amazon. the weekend scanning family photos with grandpa. its wearing on me
I'm not saying that AI is theft, but the impact on society could be compared to the same as if digital theft was allowed from one day to another. edit: but funny enough big companies pushing for AI to cut costs will eventually also suffer, then they will try to take monopoly on AI.
You'll still be able to do those same fun steps in creating film. AI is just another magnificient instrument added to the giant toolset.
@@aymanbenbaha yes, of course those responsible for creating the work in the dataset can continue making videos, they will be unknowingly be used for new datasets GEN-6 or whatever. this doesn't remove my concerns about data secrecy and shameless capitalism and the normalization of exploiting intimate personal moments from people without their consent. I'm not anti-AI and I use it for years and train models of my own from time to time but just not so excited about perfect video generations that I can overlook how its made and sold
@@Champignon1000 yeah it can be theft for sure. in certain contexts it's mathematically calculated theft to avoid copyright and leapfrog the process at the direct expense of those who didn't. sounds theivy right lol
we're done
You guys really need to either adjust the prices, or allow for hundreds of generations in a month. You guys have priced out a lot of people.
I can only get cartoon videos... How do you get the realistic scenes and people?
Price is way too high. I wonder if we’ll get image to video… if not, will never be worth it for lots of people.
its so expensive, 62 seconds of generations PER MONTH... thats literally years to create one clip if you retry a few dozen times, what the hell
Unlimited generations on the $95/mo plan
@@handfuloflight I imagine thats worth it, but its alot! not around the price for people that just wanna mess around, 60 seconds isnt good enough
@@incription ikr, I was thinking the same exact thing, I need at least to generate 5-20 minutes of content to get a feel for it. Needs to at least give 10x the amount of generations...*sighs*
does it have image to video. that's what i'm waiting for
It's impressive, but you don't give any possibility to test it and know if we will get performance out of it.
Gen-2 uses 5 credits per second of video generated.
Gen-3 Alpha uses 10 credits per second of video generated.
I have 10 credits in my account, how is it possible that I can do something viable in 1 second? How am I going to practice with one second?
The basic plan is $15 a month, that's 62 seconds of generations per month.
And that's assuming that it gives you what you ask for the first time, because if he doesn't give you what you want, you'll blow your credits just by learning.
From my point of view, I am not going to pay a subscription if I can't even know how to take advantage of it.
It is a pity. You should consider giving carte blanche for a while to people who create content using artificial intelligence, because there are many platforms that offer a good trial period, and a new one is discovered every day.
You are great, but I can't test you, therefore I don't know what results you can give me.
If I have a trial period and it works for me, I will probably subscribe, but not otherwise.
Honestly, your application is the one that I have tried the least,and I have not visited you again.
But you are great, and thanks for this video!
all the best 👍💪