This Free AI Video Generator is Wild!

A brand new tool was just made available for text to video generation and the results have been absolutely wild so far. I think we have a brand new standard in what text to video can actually do. In this video, I'm going to break down some of the coolest generations that I've come across and show you how you can use a tool like this yourself, totally for free. This video is going to be quicker than most of my videos. But I do want to show you a couple other things before I get into this new text to video model. So let's dig in. So here's some new research called Panohead, which essentially allows you to make a 3D head based on just a single image. You can see these are 3D generated heads right here where it's rotating around the head. But if we scroll down, you can see in this example there's a picture of the rock. And what it turns it into is this sort of 3D, you know, rotatable version of the rock. Now, his head shape is kind of not perfect because it's sort of guessing. And what we're seeing here with this little animation is essentially a gan. You might have heard the term GAN before. It stands for generative adversarial network. And basically what it's doing is it's looking at the initial picture, it's making a picture and saying, does this picture look like this one? No. Okay, how about this one? No. How about this one? And then it keeps going back and forth until it gets as close as it possibly can to the original picture. Now that's a total oversimplification, but once it figures out a close enough image, then it tries to guess essentially all angles of that image. Here's an example that I came across on Twitter from Ackman's where they're trying to generate the image on the left. Once the image on the left is finally generated and looks close enough, then it generates the 3D version of it. Here's another one. It's generating that image on the left and you end up with this 3D image here. And here's another one. I think this one's interesting because you could tell that her hair is sort of supposed to be in a bun, but then when it guesses what the hair is supposed to look like behind, it doesn't quite look like that bun. And then again, here's another example. Now this Pano head is open source research that is available on GitHub. You can actually find it here at this URL. I'll make sure it's linked below in the description of course. However, if you do to use it it does say here under the requirements 1 to 8 high end Nvidia GPUs. So if you have an RTX 3090 or something similar at home, you probably could use this on your home computer, but I have a feeling it might take a while. I did a search on both hugging face and replicate.com and have not found a publicly available cloud version of this yet, but I imagine it's only a matter of time before we'll be able to use it in one of those cloud environments. Now this other research that I recently came across that I wanted to show off real quick was called Motion GPT motion as foreign language. You can see it essentially generates text to motion. Can you show me a person who is practicing karate kicks and it generates this text to motion of somebody generating karate kicks? It's also capable of motion to text. Explain the motion demonstrated on and then in the little video that we're seeing over on the right here in English, so you see the character walking around and then the response from the computer. A person walks in a semicircular pattern. Tiptoeing could also predict the next movements from what it sees. And you can see it's trying to predict the next movements. Down in the bottom left here we can see some more examples here. A person is walking forwards, but stumbles and steps back, then carries on forward. You can see they're walking forward, they kind of stumble a little bit, then continue. A person moves their hands back and forth as if using a broom. So this is pretty interesting research that's coming out. Unfortunately, it doesn't seem like we have access to it yet. If we look at their GitHub repository, it explains what it does and you can read a little bit more about how it work, but there's not much information about it just yet, so we don't actually know when we'll be able to use it ourselves. Now let's talk about text to video. Previously, if you wanted to do text to video, you'd have to use something like Runway ML. Personally, I am a huge fan of Runway ML. It does generate some pretty good generations, although my fish seems to have tails on both sides of it. And generating videos does cost credits, which if you're generating a decent amount of videos, you will eat through credits very quickly. The costs do add up fairly quick. Your other option other than Gen 2 has been something like Model Scope, which you can use for free on Hugging Face. And you can see here's one that I just quickly generated that says a monkey riding roller skates and we get something that looks like this. But if you remember from past videos where we've talked about Model Scope, they all seem to have this Shutterstock watermark across pretty much every video, because clearly it was trained on Shutterstock data. But now, recently we've been given access to Zero Scope, and this one is actually available for free over on Hugging Face. You can find it at this URL here. I'll make sure it's linked below. Now, it still makes fairly short generations, but as you can see from this one, it doesn't have the watermark and it actually feels slightly more coherent than what we were seeing from Model Scope. I will warn you however, though, if you do want to generate something with the free version of Xeroscope over on Hugging Face, the generation time can usually be fairly long. And if you're using it at peak hours, it may just not work at all and tell you that it's too busy. A monkey on roller skates Submit oh, something went wrong. The application is too busy. Keep trying. So apparently there's too many people using it right now and I can't generate another video. However, you can duplicate the space if you'd like to, and the recommended hardware is using an Nvidia A10G, which will cost you about $3.15 per hour. However, using this method, it only takes about one minute to generate a video. Less than a minute, maybe 55 seconds or so to generate a video for $3.15 per hour. You could theoretically generate anywhere from 50 to 60 videos if you were really fast at prompting. I duplicated the space here myself, so we'll play with this in a second. But before we do, I want to show you some of the cooler generations that I've come across so far, just to show you what this is capable of. So this video is from Pharma Psychotic here, and it's this really cool generation of this, like robot cat that has lasers or guns or something. And I just love how this one came out. It's so cool. I I mean, you're not getting something like this out of Model Scope. Here's another one I came across from Spencer Stirling. It's like this weird underwater creature sort of scenario, but the colors and the sort of definition, the quality to it just seems to be so much better than what we were getting out of Model Scope. And quite honestly, although I do love Runway and I love all their suite of tools and I love what you can get out of gen 2, I think what we're getting out of zero scope right now actually a little bit better. Oh, this is probably not the best example because this is a bunch of creepy looking sea monsters or something. Here's another one that I really enjoyed that I came across from Vanya. But this is like this celebration with fireworks and everybody cheering and then we've got this sort of psychedelic visuals that they just blow me away. I love the colors and I love the definition of these videos. Here's one I came across from Lyle. I can actually play the music in this one because the music was generated with music Gen. But this one has that sort of painterly style. They almost look like they can be like Vincent Van Gogh paintings that came to life. And this was generated with Xeroscope. Here's one that I came across from Rupe Renisto. I'm sorry if I mispronounced your name. And I think it's supposed to be Jerry Seinfeld. It's so easy to get pot. Is it strange how socks go into the washing machine as a pair and come out single? Men and women often seem like they're from different climates. I just think it's hilarious. Obviously these aren't fooling anybody into thinking this is an actual real video of people and the images sort of blur together inside of the video. But there's something just fun and interesting about watching these videos, knowing that AI generated these videos and then getting this, you know, weird kind of borderline creepy result as you watch it. Here's another one that I came across from 3deal over on Reddit. Instead of the AI video subreddit, you can see we've got different characters walking. You've got the knight, you've got the soldiers, you've got a robot. Got a different style robot. It's just really cool. If you're wondering how they got these videos that are longer than three seconds, they're just generating a bunch of different videos and pushing them together. This one, they may have even just generated one video of, you know, a monkey walking or a person walking and then run it through something like Gen1 to change what the image looks like. I'm not totally sure how they achieved this effect. And here's Xeroscope. Here's what it looks like when you use it. Instead of hugging face. I did duplicate the space so that I could generate whatever I want and do it fairly quickly. So I generated a monkey on roller skates. Here's this one's version of a monkey on roller skates. A little bit more cartoony. It didn't try to go for that super realism. Let's try to recreate some of the other ones we saw earlier. Colorful underwater sea life. You can see it's estimating about 53 seconds to generate this video. And here's what we got out of that. If you remember my earlier gen 2 generation, that one looked a little bit less like a fish than this one, if I'm being honest. Let's do a swimming aqua octopus in a vibrant blue ocean. This one's going to take about 51 seconds. And here's what we get out of that one. Not bad. I mean, you definitely know what it is. Elon Musk wrestling with Mark Zuckerberg. And here's what that looks like. I don't know what's going on now. I actually took a whole bunch of generations of Elon Musk and Mark Zuckerberg fighting and this was the result. And that's exactly how I imagine it going down too. So that's called Xeroscope. Again, you can use it 100% for free on Hugging Face if you are patient. I haven't personally found a way to install it on your own computer and run it locally, although that's not to say there isn't a way. I just personally haven't found it yet. So right now your best option is either using it for free on Hugging Face but waiting or duplicating the space and you can generate about a video a minute. And there you have it. There's a new text to video AI tool, Zero Scope, that's available for anybody to use right now. You can create some fun videos. I just showed you a teeny tiny handful of what I've found on Twitter. But a lot of these videos are kind of going viral right now, so if you look you'll probably find a lot more. They're real fun to create, they're real easy to create. Whatever you can imagine, you can generate a funky looking video of. Hopefully you enjoyed this video. If you like Nerd not about this stuff, check out FutureTools IE where I curate all the latest tools and news that I come across. And also if you haven't already, join the free newsletter. I send it out every Friday. It's the TLDR of everything you missed both tools and news of AI for the week. You can find it all over at FutureTools IO. And if you haven't already, maybe consider giving this video a thumbs up and a subscribe and a bell and all of the stuff because that will help me with the algorithm. And also it'll make sure you see more videos like this in your newsfeed. Thank you so much for tuning in. I really, really appreciate you. I'll see you guys in the next video. Bye. Bye.