Runways Text To Video "GEN 4" Actually Changes The Industry!

15:42

Runways Text To Video "GEN 4" Actually Changes The Industry!

TheAIGRID 31.03.2025 15 579 просмотров 383 лайков

Machine-readable: Markdown · JSON API · Site index

Смотреть на YouTube

Поделиться Telegram VK Бот

Транскрипт Скачать .md

Анализ с AI

Описание видео

Join my AI Academy - https://www.skool.com/postagiprepardness 🐤 Follow Me on Twitter https://twitter.com/TheAiGrid 🌐 Checkout My website - https://theaigrid.com/ Links From Todays Video: https://x.com/iamneubert/status/1906719031018582048 00:00 – Intro: Why Gen-4 Might Be the Real Tipping Point 00:30 – Not Just Hype: Key Filmmaking Features Unlocked 01:19 – 🧍‍♂️ Character Consistency: The Foundation of Storytelling 02:11 – Why Previous Models Failed with Character Recall 02:34 – Consistent Characters = Real Narrative Possibilities 03:19 – 🔁 Object Consistency Explained 04:01 – Why Inconsistencies Kill Realism in AI Video 05:19 – Human Perception & The Importance of Visual Stability 06:13 – Full-Scene Consistency = Cinematic AI 06:19 – 🎥 Directing Characters Across Scenes 07:04 – Real-World Object Control in AI Scenes 08:21 – 🧰 Real Use Case: Putting Objects in Any Scene with Consistency 08:41 – 🔥 AI VFX: The Most Underrated Gamechanger 09:05 – Traditional VFX vs AI VFX (Cost & Time Breakdown) 10:29 – Creative Freedom: Reshoots, Lighting, and Experimentation 11:30 – 🎬 Multi-Angle Generation: AI Becomes a Real Director 12:33 – 🔬 Realistic Physics in Gen-4 13:46 – Benchmarking Physics IQ (and Why It Matters) 14:30 – Model Comparisons: Gen-4 vs Others 15:06 – Final Thoughts & My Favorite Clip of the Week Welcome to my channel where i bring you the latest breakthroughs in AI. From deep learning to robotics, i cover it all. My videos offer valuable insights and perspectives that will expand your knowledge and understanding of this rapidly evolving field. Be sure to subscribe and stay updated on my latest videos. Was there anything i missed? (For Business Enquiries) contact@theaigrid.com Music Used LEMMiNO - Cipher https://www.youtube.com/watch?v=b0q5PR1xpA0 CC BY-SA 4.0 LEMMiNO - Encounters https://www.youtube.com/watch?v=xdwWCl_5x2s #LLM #Largelanguagemodel #chatgpt #AI #ArtificialIntelligence #MachineLearning #DeepLearning #NeuralNetworks #Robotics #DataScience

Оглавление (20 сегментов)

Intro: Why Gen-4 Might Be the Real Tipping Point

so Runway Gen 4 is here and with it actually brings a suite of different innovations that makes this one actually useful and I do think that we just reached probably this week the Tipping Point in terms of AI creativity because whilst yes we've seen products like Sora and other video models this one actually has some key features that aren't just video generation but some ones that you actually need for film making that set it above the r so in this video I'll

Not Just Hype: Key Filmmaking Features Unlocked

dive deep into what makes this specific runway model Gen 4 actually so much better than the competition and why this model isn't actually just hype but rather a fundamental tool that many creatives are going to start using I mean taking a look at some of the footage here you can see that a lot of these scenes are completely beautiful and there seems to be really consistent scenes going on here and I'm not seeing so much defamation that I'm used to seeing in a variety of other clip so let's dive into what makes this really impressive because honestly I think a lot of people are going to miss out on why this model truly changed the game so one of the things that really changed the game for me and looking at this as someone who previously I wouldn't say I worked in film but I've definitely made a few short films before

🧍‍♂️ Character Consistency: The Foundation of Storytelling

consistent characters are a key gripe in terms of being able to actually make something worthwhile and Runway Gen 4 actually introduces consistent characters when you're trying to create a video now if you aren't familiar with why this is such a big thing it's because if you creating a short film a movie something like an advert you need to be able to have a consistent character that the viewers can consistently recognize and this is something that previously we weren't able to get one of the initial drawbacks from Sora was the fact that when the artists were working with it they had to prompt the model several times I think around 20 times in order to get a usable shot simply because every single time they prompted the model it was a different person every single time so they were

Why Previous Models Failed with Character Recall

basically just using the generative model see if they can get to a similar person every single time and that was something that was you know frustrating to them so think about that once again there's a huge time save there and you know this is the foundation of Storytelling in any story we always follow what the character is doing isn't just you know some cool scenes that are just animated to look really nice it's

Consistent Characters = Real Narrative Possibilities

always following a you know character going on that Journey finding out what they're up to you know their limitations whatever they explore the flashbacks the future versions I mean it's all part of the storying process and this is something that is really important now not only that it's actually going to save time you can reprompt the same character with highly specific descriptions you can use image anchors for Star references I mean this is going to be something that has Game Changer because now you're probably going to see people actually using this in their workflows because now you can generate something that people could actually use beforehand it was maybe for the off scene in a creative film but now this is something that I think a lot more agencies and individuals are going to start experimenting with and it's the

🔁 Object Consistency Explained

first model ever that we're putting out that achieves World consistency that means that you can create consistent worlds with consistent environments objects locations and characters when you can do that you can start to tell longer form narrative content with actual continuity you can generate the same characters the same objects the same locations across different scenarios so you can block your scenes and tell your stories with intention over and over again it's been very fun to design and develop Gen 4 for the last couple of months and see what the model is really capable of doing we believe that the best demonstration of Gen 4 as a creative tool is in the stories that can be told using the model we also put together a series of short films and experiments to Showcase what you're capable of doing with Gen 4 now

Why Inconsistencies Kill Realism in AI Video

not only that but there is also object consistency so another thing like I said before is not only the character consistency but object consistency whether it be a car or a solid object that you need to throw around this is going to now be consistent in the film once again like I said before this is something that you know we really do need if we want to push the boundaries and having a video generation model that is able to keep that object in all videos the same is actually pretty difficult and I'm surprised they managed to do it so you know one of the things that you know they really struggled with as well like when you're using these models I know some people just watch the tools they don't really use them but one of the things that you'll quickly realize is that without you know character consistency and object consistency things tend to lose that fun element very quickly because you realize every time you kick generate everything is just a bit random so you know when people objects and environments change every single frame it instantly starts to look fake you know humans have a really good ability to be able to tell what is real and what is fake and if you're seeing an object morph and change subtly in every single new frame you're going to realize that you know it is AI generated so you know

Human Perception & The Importance of Visual Stability

a cup turning into a different Cup every second or a car changing shapes or colors mid shot you know it just destroys the realism and humans are hard hardwire to notice those inconsistencies you know in live action animation we expect every single character every single prop to stay the same unless something actually causes them to change so if we can't have objects that can't maintain consistency then you know you can't really use that model so this is something that you know once again is super important for the development of this and this is why I said a lot of people are going to miss this update they're going to realize that well this is something that you know you definitely needed but like I said before a lot of people won't miss this because it's not going to break headlines but underneath you'll just start to realize that we're getting a lot more interesting videos and a lot of these you know consistent characters are starting to look a lot more realism you might be thinking wow when did we actually have that week or that day when

Full-Scene Consistency = Cinematic AI

we got consistent characters object consistency with a model that really

🎥 Directing Characters Across Scenes

understands the world and with Gen 4 you can actually direct the subject across the scene so you'll see we give the skunk two marks so in this case we wanted the skunk to go first in this side and then return make it feel like he's looking for something and you can see that throughout the scene is the same character with the same light with the same mood with the same condition and here we're introducing the character the protagonist right yeah again like all great animation you see so much expressiveness in the way that not only the characters are designed with the model but also moving within these scenes okay so what we just saw was a really great example of character consistency bringing those same characters across different scenes different lighting conditions with different emotions and actions being directed but you can also just take objects from The Real World so

Real-World Object Control in AI Scenes

we have a little toy here we're going to take a photo of it and then you can basically bring that photo of this toy into Gen 4 and then bring it into any environment you want so um I'm going to do a kind of little demo here I'm just going to take a picture of um this particular object that we have here again just from my phone so you'll see it come here um I'll do that and then what I'll do is I'll take that picture um and I'll run it through a set of different en enironment so now that we have that picture I'll just drag that into Gen 4 over here as my reference and I'll also use this photo of New York that I took a couple days ago and now I'm going to type the kind of composition that I want so uh in this case I want this wooden toy maybe leaning or next to the sidewalk of the street in New York City now J 4 is combining those references and there we go we have four initial images that we can browse and select I kind of like this one so it's animated and maybe one that I want people kind of like coming in front of the toy and now you can see how we get the motion and animation that are required both with the toy and the city uh uses a reference and of course you can do this with any location so we can take the toy um to perhaps the mountain we can bury it on the desert basically you can do with it whatever you want now this arguably might be the

🧰 Real Use Case: Putting Objects in Any Scene with Consistency

biggest change and I think a lot of people once again won't realize this so what this is AI V effect and what they're actually calling this is GV effect so basically if you don't understand what visual effects is this is the after effects on certain scenes that you otherwise would have thought is real now VFX is often so good

🔥 AI VFX: The Most Underrated Gamechanger

that you won't realize what you're watching actually has VFX on top of it you know as the VFX artists say you know good VFX you can't really see and the thing about VFX though is that VFX is probably one of the most timec consuming things of any you know film making scenes that you have to do you know in these big Hollywood films VFX not only

Traditional VFX vs AI VFX (Cost & Time Breakdown)

does it take time it often costs millions of dollars and they're paying these Animation Studios you know millions of dollars and you know these Animation Studios they are essentially paying millions of dollars in server Farms to rent out gpus so that they can render all of the stuff as quickly as possible so VFX usually it actually requires you know huge teams working frame by frame and you know with AI complex tasks like rotoscoping object removal motion tracking all of this stuff can now be automated like what used to be tedious manual labor can be done in minutes or in real time so instead of you know 40 minutes cleaning up a shot you know Runway can handle it near instantly and like I said before one of the biggest things that you know these companies are focused on is the cost maybe you're a agency or a short film studio and you don't want to have to spend you know millions of dollars on VFX you can use a tool like Runway to enhance that production level quality I'm sure some of you have probably seen a TV show and you've been thinking wow the visual effects on that shot is just absolutely awful and guys it's not because the VFX artists are bad it's often time because number one they may have run out of time because like I said before it is time consuming because you need to get it to look real and number two you know they may not have enough you know budget to be able to pay good enough people so that they can actually get the VFX that looks really realistic

Creative Freedom: Reshoots, Lighting, and Experimentation

maybe they're just paying someone really cheaply and of course if you're paying someone cheaply and quickly to get it done it's never going to come out well so you know this is really also you know good because if this you know is using AI if using a generative model you can actually quickly test different looks swap out the environments adjust the lighting and you won't have to re-shoot rebuild sets in 3D and this makes the creative process actually much fun much faster and more experimental so if you wanted to go from a sunny place to being on fire you could do that so I think you know in a few years this is probably going to be something that is probably you know state-ofthe-art I do think in a few years it will be normal to have aiv effects there I'm probably going to have a clip of them actually talking about this but this is definitely one of the biggest things because this is going to disrupt the industry in a way that we've never seen provided that you know you can actually upscale it and get the quality that you need because like I've said this is something that is not only is it very expensive it is also very timec consuming now there's also

🎬 Multi-Angle Generation: AI Becomes a Real Director

something else okay and I think this was you know pretty Innovative they allow you to have multiple shot angles so let's say you wanted a shot of a person you wanted a side shot you wanted a shot from above Bird's ey view this is really important for creating you know just Dynamic scenes that have different camera angles because you have so many camera angles that you just simply won't pick up on like if you're just watching a TV show you don't realize how many different you know Dynamic cam camera angles there are like you know they got the close-up shot You' got the shot from behind you've got like an overhead view you've got bird's eye view you've got you know just many different shots that just really create a dynamic scene and now that you can actually do this in Runway this is going to allow a completely different level of creative expression and of course with character consistency I don't think people are going to realize what they're watching his AI often times because it's going to be so realistic now one of the things I do just want to really get on to is of course the physics so they talk about the fact that

🔬 Realistic Physics in Gen-4

you know they've managed to upgrade the physics simulation and I believe them because I remember watching the previous edition of this model and if you actually took a look at some of the fluid simulations they did look relatively realistic compared to what I've seen from other models so I don't know what training process they used what kind of data they used maybe they just you know trained on a bunch of different fluid simulations is quite possible that they did that or fire simulations or you know water simulations it probably they probably did do something like that so they could probably you know have the best effects but this is something that is really important if we're going to have something that looks realistic and physics is really hard like I don't want you guys to underestimate how hard it is to make a model really have the correct physics because it shows us that the model is grounded in reality and often times we realize that these models are not now I'm not going to dive into all the research papers that talk about you know physics and this and that but there is one paper that I don't think most people have seen that I would like to see more models get benchmarked on so this is of course physics IQ and they talk about do generative video models understand physical principles

Benchmarking Physics IQ (and Why It Matters)

and the last time this was tested You Know state-of Thee out models got around 20% so I would like to see what this model gets on this because basically what they do is they take a video of something real happening in the real world in the real physical world and then they put an image and they explain what happens and then they see if the video models can generate the actual event occurring like it would in real life and 20% of the time those previous models do so I would like to see how this model does perform on this Benchmark now if you want to see how this model compares to other models thanks to curious Refuge we get to see how it compares to other video generation models Lumar Ray 2 actually surprisingly looks good here I haven't had you know much success with that cling doesn't look that great here but

Model Comparisons: Gen-4 vs Others

you know I think this is just one scenario so I will have to say that you know with this model definitely do test it out yourself of course there's going to be different you know scenes and different styles that may be more effective than others you know we do see that Gen 4 here compared to the others it does look the best but I think that if you're probably using Gen 4 you're definitely going to be some kind of filmmaker because it seems like that is what you know it was tailored towards that professional highgrade level film making not that kind of wacky crazy style like you know vo cling and all the all those kind of you know AI image

Final Thoughts & My Favorite Clip of the Week

generators of course I'll definitely be testing the mod out when it gets rolled out but that's something that I will be looking at now I did want to showcase to you all my favorite clip right here this was something I saw on Twitter I'll leave a link to this because this was just super creative and I don't know why but this just captivated me so I thought I would share this because it just goes to show what you can do when prompt the AI well I'm not sure who tweeted this but I will leave a link to it but overall I do think that this is probably going to change things a lot it definitely was a crazy week in Ai and if you enjoyed this video let me know what you enjoyed and I'll see you guys in the next

Другие видео автора — TheAIGRID

Ctrl+V

Экстракт Знаний в Telegram

Экстракты и дистилляты из лучших YouTube-каналов — сразу после публикации.

Подписаться

Лучшие методички за неделю — каждый понедельник