This AI Video Site has EVERYTHING (and more!) | Pollo AI

Hello everyone. Welcome back to the Matvid Pro AI YouTube channel. Today we're talking about an AI video generation website and hey, it wouldn't be the first time that I've talked about an AI video generation website because there are so many out there. It's actually a super competitive space. So what sets this one apart from the rest? What makes it special and why are we talking about it today? Well, here is the website in Take a Glance. You might not have heard of it, but this is Polo AI Pollo AI. Not exactly sure how it's pronounced. We've got an Explore tab right off the bat which insinuates some sort of social media esque features which I love to see in websites like this. And we'll get into why we've got image to video, the most popular version or popular way to generate AI videos from most generators. We've got consistent character video which I'm really interested in exploring because this has been a challenge for most AI video generation websites. Obviously your traditional text to video, they've got video to video as well, some templates. Templates, a general AI tools section and then finally where you can see all your creations. So what sets it apart? First of all, we have a plethora of features right out of the box. Some features like consistent character video is something that you don't really see again from many websites. So they have those extra features but also seem to have the features that generally you need the most, like image to video. And you know what, why don't we start there? Typically when you're generating AI video, image to video is what you want to go with. It's usually pretty quick and high quality because you're starting off with an input image to kick things off which you can devise from, you know, any picture on the web or obviously using an AI image generator to generate that initial frame which is then going to be converted into a video. Right now one thing you're going to notice right up at the top is that there is a Pollo 1.5 model. This is their own in house AI video generation model which you can try out all you want. I believe the sample video over here is from that specific model and it does all right. I think in comparison to the rest of the competition it looks pretty state of the art, but nothing totally mind blowing like we've seen lately with something like Google's VO2. Now if we click this dropdown we can see quite a few other models to pick from, which is really awesome. While this company, Polio Polo AI does have their own model and it seems again to be pretty state of the art. We've also got plenty of other models like cling, which is a super popular choice. We've got VDO AI which is an up and coming, often overlooked model which is getting a 2.0 upgrade pretty soon by the way, and also supports fast video generation at only 45 seconds on average to generate or so. We've also got Runway's Turbo model in here which is another pretty decent model in my opinion. I've spent a lot of time with that one specifically. Looks like that only takes about 30 seconds to generate. They've also got Hailu AI which is a very high quality generation model at about 3 minute generation time and then finally Luma AI at the bottom here at 45 seconds. So a really nice mixed bag of both high quality larger models and also faster Turbo models for whatever given use case you want to generate with. And this alone is something that you don't see on many video generator sites. There are others that do have access to multiple models like this one, but the fact that you're getting all of these reputable high quality models in addition to Puyo's own in house model is just nice to see. You know what, let's try that in house model. It looks like if we want to try it, obviously we're going to have to spend some credits here. It's a credit based system on this website, so yeah, they've got plans. A lot of people want to see free AI video generation. Most websites don't really have free video generation, but I will tell you, you do get some free credits here. If you go up to the top and click, check in and claim two credits each day you log in, you can get more and more credits. Obviously this is some sort of a promotion. Still though, I do like to see the fact that they are willing to benefit heavy users of the platform. So let's check in. Two credits. There we go. Does not seem to be enough credits though to actually generate a video here with Puyo 1.5. So let's take a look at the plans. They've got some yearly plans, but I'm going to go with monthly. Looks like the cheapest they have is 15 bucks a month for light, about 30 standard videos. It's pretty on par for pricing in comparison to competition. Two simultaneous videos at once, you get both standard and professional mode, faster generation speed, no watermarks on your outputs, camera movement and priority support. Alright, good to see some of these claims. We aren't going to be able to validate today, such as priority support. So we'll just hope that the support's good. Anyways, if you want to double up that price, you move from 30 standard videos to 80 standard videos, which is a pretty big increase. Three tasks simultaneously and then pretty much all the same features. Otherwise, I'm glad to see that they aren't locking things like professional mode or faster generation speed down under the larger Pro plan in comparison to the Lite. I like that you pretty much have the same plan overall. And also this one is scalable. So if you really like this website and want to do a lot of generating, you can obviously bump your price all the way up to 900 bucks a month if you're trying to generate 5,000 videos a month or something crazy like like that. So it's nice to see that it's scalable. Anyways, let's stick with the light plan. Hey, there we go. And it looks like you can also get more credits by referring friends. Oh, they just give you a nice invite code. Well, maybe I'll just toss the invite code down in the in the description below. All right, now we've got some credits to play around with. Let's give this thing a nice test. I'm always interested to see when we have smaller in house models like this created by the Puyo team here. Let's see how it really stacks up. I'm going to go ahead and upload this image right here. 11 character in the middle of jumping rope. So I want obviously for the AI to continue this. It looks like they do have a cropping feature as well, which is pretty cool. But we're obviously going to leave that on a standard 16x9. But yeah, if you wanted to zoom in or something. They do have some light cropping. Professional 3D animation. Anthropomorphic lemon character jumps rope in a field, swinging rope, animation, jumping character. I think this will be a decent prompt. They also, by the way, do have the option of adding an end frame. Always good to see that. Not going to do an end frame for this particular video, by the way. We can also make it publicly visible. And again, this insinuates some sort of a social media aspect which we'll talk about later. They've also got copy protection. Once this option is enabled, other Puyo AI users won't be able to see your uploaded image, video and prompt for this generation in the community. Well, some people definitely like stuff like that. I'm not too concerned with it personally. So let's just click create since we can do two simultaneous generations, let's go ahead and switch models to hell u AI. Oh my God. 50 credits to generate with hail u AI. Never mind. Let's try cling 1.6. All right. That's only 20 credits. I like the fact though, that it kept my image and my prompt in there when switching models. It's subtle user experiences like that that make all the difference in my opinion on websites like this. The fact that I don't have to copy and paste the prompt and put the image in again. I can just switch models and generate. We'll do standard mode. Now. I do want to talk about a little bit about the whole credits situation on a lot of websites. Like, let's say this was just Pollo's, you know, in house model. This was the only model you could use if they had a credit system. I might be a little bit more apprehensive and abrasive to something like that, but since we have all of these different models and obviously they're pinging different APIs, these models are going to cost different amounts each to run. And honestly, with that, it makes sense that they would have something like a credit system. Poyo is pretty cheap to run, but CLING is not so much. Now, while we wait for these videos to generate, it's also worth mentioning, folks, that Puyo AI did sponsor today's video. Sponsors are super important for this channel. It's literally the reason I could actually pursue this channel full time. If it wasn't for sponsors, I wouldn't make enough for that. I'd have to have some other job. I want you guys to know though, that even though this video is sponsored, it's all my own opinions here. And as we go through exploring this website, if I see something I'm not a huge fan of, I'm going to let you guys know. Our cling generation came out first. This is cling 1.6. Now obviously guys, the whole jumping rope is going to be pretty difficult for these video generators. That's why I chose it. Cling 1.6 struggles a little bit here does not produce the full jumping rope looping back and forth like we expect more of just holding like two little strings in each hand and jumping around with them. Other than that though, I think the character looks decent. A little bit of morphing on the face, but the flowers swaying in the wind. I do like that. Now, how did Puyo AI's native model do for this prompt? Looks like not too well in comparison to cling. Both of them didn't do great, but clings is clearly better. You can see it's trying to get there. Again, this is a particularly difficult prompt. He's sort of just kind of doing some movements, and then eventually everything kind of mushes up and turns into AI stew, so to speak. But this is why we do simultaneous generations. This is why we have access to multiple models in the first place. We'll take our better video here by cling, and we will upscale it. Let's go ahead and do a 2x upscale. They've got different scaling modes as well. Standard face denoising. All pretty cool. We'll just use standard mode and generate it. Looks also, by the way, like you can upload any video up to 200 megabytes in here and just upscale it. So nice to see that you could use pretty much any video you want with this upscaling tool. Oh, and as you can see, it actually changed us down to the extra AI Tools section where we can see Video upscaler. Video Enhancer Filters. Dance Generator. Motion Brush. This looks like basically a collection of a bunch of different tools that don't exactly fit in any of these other major categories. Still, though, lots of features on here, man. If you're going to have an AI website where you're charging for monthly plans that give you credits, this is the kind of thing we expect. We expect you to have face enhancers denoising video tools, video to video filters, for example. Extra features. Anyways, let's now take a look at consistent characters. Our upscales are already done, by the way. That looks quite a bit better. Pretty good upscaling. I don't know if this is their own in house upscaler, but it looks pretty good. So for consistent character video, I'm not sure what they're doing or what models they're using under the hood to make this happen, but let's give it a shot. Image one, image two, and then image three. I feel like the more images we have in here, the more consistent our character is going to be across the video. Let's prompt some images for our consistent story. We'll have the character looking at a pizza restaurant from the outside. Now let's have this character eating a pizza. So he obviously bought a pizza looking bloated and satisfied. Obviously, we're going to be at the mercy of Ideogram's ability to generate a consistent character. We will just have to try to pick the most similar characters possible. Start out with this guy, and then this one's pretty close. Ish. And then, you know, What? This is just way too good not to use. So we're going to use this as our third image. We'll just see what happens. Sure. All right. This is kind of a bad prompt, but he sees the pizza place. He stops in to eat the pizza, then rubs his tummy after eating it. Sure. Well, I have a long video length for this and then the motion range. I guess we'll just leave that on auto. This is going to cost a whopping 100 credits. Let's see how it does. Okay, the video is done generating and wow. I have a lot to say, actually. So take. Take a look at this, guys. Right. We're starting off, obviously with this character and he's at the pizza place. Clearly in the background we can see. Then he eats the pizza and he gets like a third arm somewhere in there, and then yum, he's happy. Obviously this generation wouldn't be usable for anything professional, but it is pretty cool. I thought it was gonna use those images more as references, I think. Obvious first one here. The first image was very much used as a good reference, but it really did all of its own generating here. At the end of the day, it kind of has combined everything together. You can see he eats the pizza. That's when he grows the third arm there. And then very, very full. And then it says, yum. So it's almost like the AI in took all of these images itself and then devised the generation based on all of them. I thought it was gonna, like, midway through, cut to this exact image, and then at the end cut to that exact image. It really did use them as reference, though, not as actual points in time throughout the video, which I think is better. Even if the video didn't come out all that great, it looks to me like this might actually be an in house consistent character model that Pollo Polo is working on. I think the first to the second actually does look pretty good. Like he eats the pizza and then things get a little bit strange right around halfway through. This sort of thing is really a challenge right now for video generation models. Like, getting an animated character to eat a pizza reasonably is difficult enough, but having a consistent character do it over time, that's difficult. I suppose all of these different images don't help too much. I kind of want to experiment more because these characters probably are too different from each other for this to work out properly. So instead, let's upload a bunch of photos of yours truly and I'll be the consistent character. Gorgeous beach scene, muscly ripped man. Yours truly Runs across beach followed by hundreds of cute dogs. Oh, man, I'm gonna blow another 100 credits on this because I'm so interested on the consistent characters. Let's see. I mean, this is obviously a much more consistent person. These are real photos of me not doing this. AI generated three different lemon characters. I wonder if this one will be a lot better. Now for this prompt up here, I will say I do kind of like the approach on the 2D animation here. It's pretty decent considering that 2D animation is kind of a weak point for a lot of video generator AIs right now, with the exception of a few, such as vdu, which we actually do have access to that model in the Puyo AI website. All right, while this generates, I'm going to take a look at the Explore tab. So this is obviously the social media aspect of all of this. We've got different people who have signed in to the website here, sharing their generations. We've got Albert Einstein shredding the guitar. That's actually a pretty cool video. It looks like we can leave likes, which is nice. I don't know if you can go through and look at all of your liked videos because you might want to, you know, save a bunch of inspiration by liking a bunch of other people's videos. You can also copy link to share it, share it to Facebook or X. Or you can actually straight up download the video with or without watermarks because he has made this one public. We went over the public settings during video generation, but I like to see. I like that you have the choice to either make it public and share it with everybody else or just leave it. Not public public. And they have a quick create similar video tab right in here, which is going to take that initial image and put his prompt in there with whatever model I assume he used. In this case, it seems to be POYO 1.5. Lots of fun generations in here, though. We've got a nice zebra. We've got a cat being squished. This is like that viral AI type video that you've seen floating around, I'm sure on the web. If I click create similar video. Yes. AI Squish effect. So this is one of their templates that they have. They've got an exploding effect. Melting venom transformation. Okay. I've seen a lot of these actually floating around TikTok, which is actually going away anyways. The AI Hug AI Joker transformation is kind of cool. They've got a lot of fun templates on here. This is fun. If you're. You Know new to AI video generation and just want to mess around and share some fun things with your friends and family. All right. Pretty cool. That's. That's what I thought. I thought that was a template. You can see a lot of these though are going to be using some of those tools and templates. Like this is I think the anime video conversion AI that they have in the AI Tools section. I like social media aspects though. I like stuff like this because it does allow you to save and actually get ideas from other creators on what you might want to generate. Obviously. Although a lot of these are sort of those template esque style video creations. This one's pretty crazy though. We've got Pepe the Frog on a rocket. I'll leave that one alike. That one's pretty good on this. Do they have. Oh yes. Okay. They have have trending. They have brand new generations. There's mine right there. You can see that I just made. And then they have most liked of all time and well, there you go. I mean AI generated Christmas influencer folks. Some of these are genuinely cool though. I do like the skeletons with the fire. That deserves a like. That one's cool and okay. Yeah, you can see there's. There's a lot of like more promiscuous stuff going on. I think the trending tab is my favorite tab so far. But it really doesn't hurt to have features like this on your website. Again, there is real value, there is real benefit in sharing your creations with others and being able to see other people's creations and devise ideas from them. If you want to make something similar but slightly tweaked, you know, if you like the anime style template, really easy to do. They even show you the original here that they made. Okay, my consistent character video has been generated. Take a look guys. It definitely looks like me if I was absolutely ripped and shredded. I mean just barely, right? I'll move my head actually. Obviously it is trying to conform to the prompt though. Running on the beach. We've got me absolutely shredded. We've got the hundreds of dogs as well. It does. It does kind of look like me. I mean it's got the hair right. The face is pretty close. Not perfect. And then we also see another clone of me running over towards the end of the video, which was pretty funny. Obviously this is far from perfect. This is a far from perfect consistent character and generated video. But it's something. It's something most AI video generators have. Have no consistent character features whatsoever and none of them are perfect. I Will tell you that right off the bat. This is fun, though. It did it. It made me the guy at hand here who is shredded, and it. It does look like me just a little bit. I like it. I also like all the dogs in the background as well. All right, this video alone leads me to believe that this is Poyo's own custom character consistency model. I think they're on to something here. Like I said, far from perfect. This is alpha early days, right? Early. I mean, a video generation tech as a whole. But you guys are getting somewhere. You're getting somewhere for sure. It was able to take those three images of me, understand what my face looks like. And while it might, like I said, be far from perfect, it's. It's definitely kind of getting the concept down. Right? I know you're watching this, guys. This right here, consistent character video. This is what I think should be a main focus for you guys. You're on to something here for sure. I don't know what the architecture under the hood is quite like a research paper on this could be pretty cool. But this is what people want. They want to be able to tell stories with consistent characters. And that's something right there that I know people are willing to pay for, especially as it becomes better. Like I said, this is still pretty rough. Also, it does cost quite a few credits to do this. This could be a great website for folks who want to start off an AI video. They want to explore it and are willing to spend, you know, at least 15 bucks to give it a shot. All right, so Poyo AI final thought right now. AI video generation is still pretty expensive. It costs a lot of compute, a lot of GPU power, and that is very much translated onto the consumer. In pretty much every video generation website you're going to come across, it's going to have plans, it's going to cost money. There are a few exceptions that we've talked about, but for the most part, I really do understand this. Pricing is right about in line with everybody else, especially while being able to access models from almost everybody else. I think they've got a really nice UI design on this website. Overall, it's easy to use. It has those subtle features again, where changing the model doesn't remove my image, it doesn't remove my prompt. That makes running and comparing different models a lot easier. All the basic settings that we might want, depending on the model we choose, like the ability to use Cling's Motion Brush. We've got all those templates, all those different AI tools. It's a nice one stop shop for AI video generation. Generation. It's competitive and in a very short amount of time. I've been following this website for a while now and it has evolved and updated quite rapidly and generation time is solid so far. Again, you know they're using APIs so some of that can be attributed to other companies, but even with Puyo's own in house models, the generation time is very reasonable. At the end of the day, I think this is a pretty solid place to start off if you want to do AI video generation, if you want to explore different models, if you want to explore again those different templates, some of the other tools they have, they've got upscaling. Not a bad place to start off here. I am most impressed by the in house models. Even though they might not be as good as some of the other players who have had maybe more time to invest or maybe even more money to invest in training their models, the fact that they're working on their own consistent character model is huge to me. That is something that again you just aren't seeing a lot right now in the space and is something that people are are dying for. So give this one a shot guys. Like I said, huge thanks to Puyo for sponsoring today's video and allowing me to do this review. This is absolutely, I think something that is at least worth checking out and again, I think it's a really solid recommendation to new users or new people in the AI video space who want to try different models, different templates and maybe explore some other people's ideas through the Explore tab. AI video is still far from perfect no matter where you go to check it out, but it is improving rapidly. I would love to see where this website is at the end of the year. Hopefully that consistent character video gets upgraded tenfold because that would put this website right at the top of the list for me personally. And let me know what you thought. Is Puyo AI a website that you're interested in checking out? Is there anything that you think I missed in this review? Are there any models that are available elsewhere that you would love to see on this website specifically? The only model I can really think of that I would like to see on here would be the LTX video one because I know that model is very cost efficient and very quick to generate and has decent quality. So maybe cheaper models on this website would be something to look into. Alright, well that's it for me. I'll see you guys in the next one and goodbye.