Transcription
This is the best free and open-source AI video generator you can use right now. And yes, this is also the most uncensored. Trust me, it's the first thing I tried. It's called Juan 2.2 by Alibaba, and this is the successor to one 2.1, which is already previously the best open-source video model out there.
So, in this video, I'm going to go over all the cool things that you can do with one 2.2. And of course, I'm also going to show you step by step how to install this so you can run it on your computer for unlimited times offline. And yes, we already have some quantized versions of 1 2.2 available. So you can run this with low VRAM. Let's jump right in.
So first, here's their official announcement page. And note that for W 2.2, it's really good at generating cinematic looking videos. So here, for example, it's really good at portraying the lighting very realistically. And then over here, it's also really good at generating this wide-angle shot and following the prompt. Same with this one. It's able to generate this tracking shot of a man as he runs down the street very well. And then same with this one. Everything just looks very cinematic and realistic. It's also great at anatomy understanding and handling high-action scenes like this.
So, here we have this hip-hop dance example. And for the most part, everything looks very coherent. And here we have a cat punching some gorillas. And you know, for older video models, it's really hard to generate these fight scenes coherently, but one 2.2 is able to do this very well. And then here is a street parkour scene. And again, it handles everything beautifully. The camera movements also look very cinematic, but honestly, I don't think their official demos showcase how good this is.
So, next, let me also show you some of my personal demos. Now, there are a few ways you can use one 2.2. One way is to use it online via their platform, and the URL is just one. Which I'll link to in the description below. You can sign up for free using Gmail or GitHub and then once you're in, you can generate pretty much unlimited generations, although it is pretty slow on the free plan. Anyways, at the bottom here is where you can start generating a video. Now, there are currently two options for you to choose from: text to video or image to video. There are also other options down here like first frame, last frame, or references. But note that if you select these ones, it only uses one 2.1.
Anyways, let's try out some texttovideo examples first. This is where you would enter a text prompt and it would generate a video based on your description. All right, so let's start with a really tricky prompt with a ton of different elements just to show you how good one 2.2 is at actually understanding your prompt. So here the prompt is a Victorian lady in a lace gown stands at an ornate vanity in a lavish bedroom adorned with gold-framed mirrors and silk draperies. The vanity holds antique bottles and a wilting bouquet of roses and a half-open jewelry box spilling pearls. We also have a parrot squawking from its cage by a fireplace. And also through the balcony doors, you can see a dinosaur grazing in the garden. Her butler somehow is wearing a superhero cape over his tuxedo. And he enters with a tray of tea. So this is a very complex prompt with a lot of elements that are not expected. So, let's see if one 2.2 can capture everything.
All right. And here's what we get. Indeed, we do have a Victorian lady in a lace gown, and she is in a lavish bedroom adorned with gold-framed mirrors and silk draperies. The vanity does hold antique perfume bottles and a half-open jewelry box. Plus, there is a parrot, although there isn't really a fireplace in this scene. And we do have a dinosaur outside which is grazing in the garden. Plus, we do have her butler walking in the scene holding a tray of tea. And he is wearing a cape over his tuxedo. So, for the most part, except for the fireplace, Juan handles this very well.
All right, here's another complicated example. A ballerina in a tutu practice spins in a sunlit studio, and the floor is somehow scattered with point shoes and sheet music. A rabbit watches a top from a grand piano. And then outside the large window, an elephant balances on a circus ball. Let's click generate. All right. And here is what we get. Indeed, we have this ballerina spinning around. And the floor is scattered with shoes and sheet music. Plus, we do have a rabbit on top of a piano. And outside, we do have an elephant balancing on a circus ball. Notice that the anatomy of the ballerina is also consistent across this entire generation. We don't get extra limbs or disappearing limbs that we've seen with older video models. So overall, a very impressive generation, especially for an open-source model.
Now, one 2.2 is also really good at camera control. So you can specify how you want the camera to move throughout the video. For example, we can create a tracking shot from behind, and it's going to be a snowboarder launching off a snowy cliff, mid-air rotation, sharp descent into fresh powder, snow dust trails behind, heart racing descent into the valley. Let's press generate and see what that gives us. And here is what we get. Indeed, this is a tracking shot of a snowboarder as he launches off a snowy cliff. And he does leave a trail of snow dust behind. And this is a descent into the valley. So again, overall, this is a very impressive generation. Notice that the physics are also very accurate. There are no deformed limbs. His head doesn't suddenly come out of his butt. Like everything is just anatomically accurate. Very impressive.
Now, speaking of anatomy, here's another tricky prompt that no other video generator has gotten correct. This includes the legendary Hyo O2 and VO3 and Cling. So, the prompt is a gymnast performs a flip on a balance beam. Let's see if 2.2 can handle this. And here's what we get. We do have a gymnast and she is trying to flip on this balance beam. And for the most part, this is anatomically correct. Like, we don't have any missing limbs or extra limbs. This is way more anatomically correct compared to 1 2.1 or even to some of the top proprietary models out there. So, this is really good.
All right, here's another tricky prompt. A cat figure skating on an ice rink with graceful leaps and spins. Let's see if it can generate this. All right, and here is what we get. Indeed, this is a cat figure skating on an ice rink. And you know, it is leaping and spinning perfectly. There is nothing anatomically wrong with this generation. Very impressive.
Now, still most video generators cannot generate fight scenes very well. So, let's try this prompt. Two men in white tuxedos fighting on a city rooftop. Rain and lightning in the background. High action, fast motion. Let's press generate. And here is what I got. As you can see, this is indeed two men in white tuxedos fighting on a rooftop. And there is rain and lightning in the background. Now, even though I specified high action and fast motion, the movements in this video are still kind of slow. It's probably because 1 2.2 is kind of trained to generate these more cinematic shots. But I'm sure there's going to be a Laura out soon which can generate faster motion. But anyways, as you can see here, again, everything is anatomically correct. They do look like they are fighting and you know, the body doesn't warp. We don't get any extra limbs. This whole fight scene looks very good.
Or let's try another similar prompt. An intense fight between a man with a cat's head and another man with a dog's head. High action, fast movements. And here is what we get. So this scene moves a lot faster. It does look like these two are fighting pretty intensely. Again, everything just looks anatomically correct. Very nice. So one 2.2 can definitely handle fight scenes and other high-action scenes.
Finally, here is your favorite prompt. Will Smith eating spaghetti. Let's see if it can handle that. All right, at least for text to video, it can't really generate existing people or celebrities like Will Smith. So, here's just a generic dude. Who knows? Maybe on planet Earth there is a Will Smith who looks like this. But obviously, this is not the Will Smith that I was looking for. But this doesn't really matter because 1 2.2 can also do image to video. So instead of text to video, if we set this to image to video, well, I can first generate an image of Will Smith eating spaghetti and then upload it over here. And then for the prompt, let's write he is eating spaghetti and then click generate. And here is our result. This is indeed Will Smith eating spaghetti. And he also eats this really realistically.
In fact, let's now test out some other imagetovideo examples. Let's see if it can generate anime. So, I'm going to upload this anime image as the first frame. And then for the prompt, I'm going to write a group of friends talking in a cafe. Let's click generate. All right. And here is what we get. This is really good. It animates all four characters. And this does look like an anime scene. The way they talk and move looks, you know, very characteristic of anime. So, one 2.2 can definitely handle anime very well.
Next, let's try a 3D Pixar example. So, I'm going to upload this image, and I'm not even going to enter any prompt. I'm just going to leave the creativity to it. Now, this is a very tricky scene with a lot of characters and elements. Let's see what it can come up with. All right. And here is what we get. This is actually a very impressive generation. How it's able to get them to move and talk and jump up and down like this. I mean, this can be a scene straight from a Pixar movie. It's really good at animating these types of photos.
Next, let's test it on an even trickier example. This is going to be like a Chinese watercolor painting. Again, I'm just going to leave the prompt empty. And let's see what it does. And here is what we get. It's pretty cool how it knows to animate only the fish, but you know, keep the rest of the background still. Very interesting.
All right. Next, let's see if it can generate a very complex dance scene. So, I'm going to upload this photo of five characters. And then for the prompt, let's write a K-pop girl group performing a cute dance. Let's see what that gives us. And here is what we get. Notice how consistent and coherent everything looks. Their faces don't really warp over time. This is super consistent. Plus, they are all doing a cute dance in sync. Very cool. This is indeed one of the highest quality image to video generators you can use right now. And I can't believe this is open source.
All right. Next, I also wanted to give it some trickier examples. So, here I'm going to upload this Minecraft looking scene which I just generated with Google's imagine. And for the prompt, I'm going to leave it empty. And let's see what it generates. All right, here is what we get. And you know, for the most part, everything again looks very good. It knows to move certain characters, but not the other blocks in the scene, which are like the buildings or the plants. This is really impressive.
All right. Similarly, here's an even trickier example of a Genshin Impact gameplay scene with a lot of health bars and numbers overlaid on this scene. Again, I'm going to leave the prompt empty. And let's see what it generates. And here is what we get. Notice that the numbers on top of the health bars are kind of messed up, but it knows to keep most of the interface, like the text on the left and right sides, consistent. And this is kind of like a fight scene. So, overall, this is not bad. I did not expect this to give me a perfect generation.
Finally, I also wanted to try something creative. So, I uploaded this image with some instructions in text boxes. Let's see if one can actually generate these elements in the video. Now, I'm just going to upload this image and not enter any prompt. Let's click generate. All right. And here is what we get. Indeed, we do have a dog walking across over here and a balloon kind of floating up, although the duration of the video is not really long enough for the balloon to float up, but there you go. Now, this method actually works really well with VO3. So, you can write instructions on your initial image, and when it generates the video, it'll actually remove the text and follow your instructions. But for one, at least for my demonstration over here, it doesn't really remove the text. although it does kind of follow my instructions. Very interesting.
So hopefully that gives you a good sense of all the amazing things that one 2.2 can do. Next, let's move on to how you can install this locally so you can run it offline for unlimited times. Now the best and most customizable way for you to run this, especially if you have low VRAM, is using Comfy UI. This is one of the most, if not the most popular open-source platforms for using image and video generators offline. If you're not familiar with Comfy UI, I already did a full tutorial on it over here. So, see this to get started. Anyways, in this video, I'm going to assume you already have Comfy UI installed.
Now, there are already a ton of different Comfy UI workflows that incorporate one 2.2. For example, you can use this one video wrapper by this goat called Kiji. In addition, Comfy UI themselves also released an official workflow for 1 2.2, and that's exactly what we're going to go over in this video. This workflow is a lot simpler and less error-prone. Now, there are actually several versions of 1 2.2 released. The smallest one is 5 billion parameters, and this is a hybrid model that can do both text to video and image to video. Now, because this is smaller, it's going to run faster, but at the sacrifice of some quality. And then we also have some 14 billion parameter versions. And for these ones, the image to video models are separate from the text to video models. So let's go over how to get all of these up and running on your computer using Comfy UI.
Thanks to Monica for sponsoring this video. Monica is an AI assistant that provides access to the best AI tools all in one place. This includes the top AI models like GPT, DeepSeek, and Gemini, plus the top image generators like Flux and Stable Diffusion, and the top video generators like Cling and High Law. This is a killer deal because you can use all of these in one platform for way cheaper than if you used all of these separately. You can use this as a browser extension on your desktop or on your mobile device. And their browser extension is context-aware. So you can interact directly with any web page you're on. You can easily get Monica to reply to emails or summarize an article. So for example, let's say I'm on this page. I can get Monica to summarize this really technical article in just one click. And here's something even cooler. You can even generate a mind map from this content, which can help you understand ideas more visually. Note that this is way more convenient than just copying everything and going to chat GPT to ask it to summarize the article. You can also use Monica to summarize YouTube videos. So, here's an example. Let's say I'm watching this YouTube video. I can get it to generate a summary or even generate a podcast in just one click. Let's try this one. It even gives you a list of highlights with the correct timestamps. Plus, if you click on this again, it's generating a mind map for you so you can understand everything more visually. With Monica, everything is just a few clicks away. But don't take my word for it. They have over 10 million users worldwide with a 4.9 star rating on the Chrome store. Try it out for free via the link in the description below. And when you're ready to upgrade, use my code AIARCH10 for a nice discount. Full details in the description below.
Let's go over how you can use this 5 billion parameter version first. Now, this is a hybrid model, so it can do both text to video and image to video using the same model. And here it says this can even fit with 8 GB of VRAM. And that's why we use Comfy UI. It's very customizable and very friendly for people with low VRAM. Anyways, before we download the workflow, it's best if we download all these models first. So, here is a nice diagram of where you should place all these models. So, I'm going to click on this one first. And this will go in my Comfy Y folder in models and then in diffusion models. Let's click save. Note that this is 9.3 GB in size. So it's going to take a while to download everything. In addition, we also need to download this 1 2.2 VA file. So let's also click on this. And this goes in Comfy UI in models and then in VAE. So I'm going to click save. And then finally, we also need this UMT5XXL FP8 scaled safe tensors file. So, let me click on this and then navigate to Comfy UI and then models and then text encoders. Now, I already have this over here, so I don't need to download this again. But notice that this is over 6 GB in size. Also note that the VAE file is 1.3 GB in size.
All right, after you've downloaded all these files, let's open up Comfy UI. Now, before doing anything else, make sure you update to the latest version of Comfy UI first. So, let's click on manager and then click update Comfy UI. And that's because we need the latest version which supports all these components for 1 2.2. So, let's click on this. And you can see for me, I already have the latest version. So, I'm going to click close and then close again. The nice thing about Comfy UI is you don't need to program all these nodes and noodles from scratch. You can just take an existing workflow and drag and drop it onto your interface. So that's exactly what we're going to do. I'm going to right-click and then click save link as. You can save this wherever you want. I'm just going to save this in my comi folder. So let's click save. And then afterwards you can just drag and drop it onto here. And that's pretty much it. It's as simple as that.
So for this first component, let's make sure to click on each of these drop downs and select the model that you just downloaded. So I'm going to click on here and select this one 2.2 safe tenters file. And then for this clip loader, let's select this UMT5 one. And then for the VAE, let's select the one 2.2 VA. And that's pretty much it. Now again, for this 5 billion parameter version, this does both text video and image to video. So, let's first try out a text video example. For the prompt, let's just do a simple one like a woman running in the city. And then over here is where you would specify the width and the height of the video. So, notice that right now it's 1280x 704. And then here's where you specify the length of the video. Notice that right now it's at 24 frames per second. So, 121 frames would be roughly 5 seconds. Now, if you've been watching my previous comi videos, then you should be quite familiar with this K sampler node. The seed is basically the starting point of your video because in theory, there could be almost an infinite number of different videos you can create with this prompt, and this is just one of them. Now, usually we just set the seed to random. And then for steps, this is basically how many iterations you want the AI to go through before outputting your final video. In general, the more steps you have, the higher quality your video will be, but it also takes longer to run, and you will get diminishing returns if you set this to too many steps. Conversely, if you set this to a lower number, then it's going to run faster, but at the sacrifice of some quality. And then over here, CFG is how literally you want the AI to follow your prompt. So, a high CFG value would get it to follow your prompt very literally, whereas a lower CFG would give it more creativity and might add more variance to your video. And then the sampler name anduler, these are basically the algorithms used to generate the video. Now, there are a ton of different algorithms you can choose from, but I just tend to leave it at the default. And that's pretty much it. Let's click run and see what that gives us. And here is what we get. Notice that for me I have a GPU with 16 GB of VRAM and this took roughly 12 minutes. But of course you can also add some additional Loras to this like self-forcing or covid. So you can decrease the number of steps and make this generate even faster. Anyways, if you look here, note that the result is kind of cropped. So in order to view this properly, you can right-click and then view this in a new tab. And here is what we get. So, it's not bad for a model with only 5 billion parameters.
Now, instead of text to video, let's also try image to video. Note that currently this node is highlighted in purple, which means it's bypassed. So, to unbypass this, you need to click on it and then press Ctrl B. Now, here's where we would upload an image to use as the first frame of the video. So, I'm going to upload this woman holding a pug. Now, because this is a vertical image, let's set the width and height to something like this. And then for the prompt, let's write she kisses the pug. And I'm going to leave the rest of the settings the same. Let's click run. All right, here is what we get. And this looks really good and super realistic. She does proceed to kiss the pug. Very nice. Again, super impressive for a 5 billion parameter model. And that's pretty much it. That is how you can use one 2.2 two for both text to video and image to video.
Now again it says here this 5 billion parameter model should work with only 8 GB of VRAM. But if you have more VRAM then you could go with this 14 billion parameter model and download this workflow instead. And of course this 14 billion parameter model is going to have higher quality and coherence compared to the 5 billion parameter model. But of course, you do need to have enough VRAM to run this. So, next I'm also going to show you how to download and run this version.
So, first we need to download both the high noise and the low-noise models. So, let's click on each of these. And this goes in comfy UI in models and then in diffusion models. And similarly for this low-noise one, we also need to click on this. And this also goes in the diffusion models folder. Note that each of these is around 14 GB in size. So, it's going to take a pretty long time to download this. And then we also need to download this 1 2.1 VAE. Notice this is 1 2.1 not 2.2. So, let's click on this. And this goes in comfy UI in models and then in VAE. Notice that I already have this over here. And then finally, we also need to download this text encoder. And then this goes in models and then text encoders. You can see I already downloaded this from the previous workflow.
All right. After downloading both the high noise and the low-noise models, next we just need to download this workflow file. So I'm going to right-click and then click save link as. And you can save this anywhere. So I'm just going to save this in my main Comfy UI folder. All right. So, back in your Comfy UI interface, all you need to do is drag and drop your downloaded workflow file onto this interface, and you should see this. We don't need to build any of this ourselves. So, again, here is where we need to select the drop-down for each model and select the one we just downloaded. So, over here, I'm going to select high noise. And then for this one, I'm going to select low noise. And then this one, I'm going to select UMT5. And then finally for this one, I'm going to select 1 2.1, not 2.2. And then here's where you would enter the positive prompt. Here's where you would enter the negative prompt. Here is where you would specify the dimensions of your video as well as the length of your video. So again, 121 frames divided by 24 frames per second is roughly 5 seconds. And then here's the K sampler. The settings are pretty much the same as the previous workflow, so I won't go over these again. And that's pretty much it. After you enter your prompt, you can click run to generate your video. So that sums up the 14 billion parameter texttovideo workflow. And the reason to use this is because the quality and coherence are going to be better than the 5 billion parameter version. But of course, you do need to have enough VRAM to run this.
Finally, let's also go over this 14 billion parameter image to video workflow. Now unfortunately for this you do need to download these additional models which are specialized for image to video. So again we have this high-noise model and this low-noise model. So let's click on each one and as before this also goes in comfy UI in models and then in diffusion models. Same with this low-noise one. And then this one 2.1 VAE is the same as the previous one. This goes in the VAE folder. And then this UMT5 is also the same as what we downloaded previously. So this goes in the text encoders folder.
All right. So after downloading both of these again, let's download this image to video workflow. So I'm going to right-click and then click save link as. And you can save this workflow file anywhere. And then back in our Comfy UI interface, all we need to do is drag and drop this 14B image to video workflow file onto our interface. And again, the nice thing is this workflow is already pre-built for you. Now, as before, for this component, make sure you click on each dropdown and select the model you downloaded. So, for example, for this one, let's select this I2V high noise. And for this one, let's select ITV low noise. For this, let's select UMT5. And then for this, let's select 1 2.1. And that's pretty much it. Here is where you would enter your positive prompt. Here is where you would enter the negative prompt. And then here is where you would upload an image to use as the starting frame of the video. And the rest of these settings are the same as the previous workflows. So that's how you get this 14 billion parameter image to video workflow up and running on your computer. All the instructions and download links for all the models are all on this official Comfy UI page. So I'll link to this in the description below.
Next up, this is super important. So, I also want to go over some ways you can speed up your generation, especially if your GPU isn't that good and if you get an out of memory error or if it takes painfully long to generate one video. Here are some hacks you can use to make this even faster. The nice thing is there are already really quantized versions of 12.2 available. So, I'm going to link to this page and this person has quantized all the 12.2 two models including the 5 billion parameter one and the 14 billion parameter one for text to video and image to video. So here I'm just going to show you this 5 billion parameter example. But you can also load these 14 billion parameter models the same way. Anyways, let me click on this and then in files and versions you can see a list of these compressed models over here. The smallest one is less than 2 GB in size. So maybe you can fit this in like 4 to 6 GB of VRAM. But honestly, with this much compression, the quality probably wouldn't be that good. So I'm just going to go ahead and download this Q8 one, which is less than 6 GB in size. So let's click download. And this goes in Comfy UI in models and then in diffusion models. You can see that I already have this over here.
All right. After downloading this GGUF model, let's open up our 5 billion parameter workflow again. And then over here, instead of this load diffusion model node, I'm actually going to click on this and delete it. And then what I'm going to do is double click anywhere and then search for gguf. And you should see this gguf loader over here. After you load this up, simply click on this dropdown and select the model that we just downloaded, which is this one. So, let me drag this back over here. And then I'm going to connect the model to this node over here. And that's pretty much it. This should load way faster than the original 5B model.
Another technique you can use to make this faster is you can decrease the resolution of the video. So maybe you don't need such a high resolution video. In that case, maybe you can set it to something like 732x 480. So those are some quick ways you can make this even faster or run this with lower VRAM.
Now, here's another hack. Notice that I believe this only works for these 14B workflows, which looks like this. Remember, this workflow has both a high-noise model and a low-noise model. The really awesome thing about 1 2.2 is it's backwards compatible with Loras from 1 2.1. Now, there are some really helpful luras that can help speed up your generation by allowing you to make videos in fewer steps. One of them is called self-forcing and you can basically download this lura and add it to your workflow. And once you do that, you can basically decrease the number of steps to like four to eight steps. So you can generate videos in less than half the time. So here is how you would load the Laura into here. So first, this depends on whether you want to do image to video or text to video. So for this workflow, let's say you are doing image to video. Well, you would have to go ahead and download this self-forcing 14B I2V model. If you want to do text to video, then you would click on this one. Anyways, let's click download. And this goes in your Comfy folder in models and then in Laura's. All right, so let's click save. Note that when you download this, the name is not called self-forcing, it's called light x2v. So these two names are interchangeable. Let's click save. And note that this is 705 megabytes in size. So it's a pretty small Laura.
Afterwards, going back here, let me just stretch this out to make it easier to follow. So, here we have the high-noise model. Here we have the low-noise model. And then this goes through this model sampling SD3 component. And then the top one goes through this K sampler. The bottom one goes through this one. We need to add a Laura in between these steps. So, let's double click anywhere on your interface and then search for Laura model. And then we can use this one, Laura Loader model only. And you should already have this because this is just built in natively in Comfy UI. And then on this dropdown, depending on what workflow you're on right now, you would select either the ITV one or the TTV one. Because this workflow is ITV, I'm going to select this one. And then let's just click on this node and copy and paste it using Ctrl + C and Ctrl + V. And then afterwards, we just need to connect this one over here. And then this one over here. And then same with this. Let's connect it over here. And then connect this over here. So basically we are loading the original model but then also adding this light X2V or this self-forcing Laura over here which is going to speed up the generation by a lot. And then note that over here it says this Laura only requires four steps at a CFG of one. And then you need an LCM sampler. So over here I tend to find that this works well with four to eight steps. So let's set this to four steps. And then for the CFG, let's set this to one. For the sampler name, I find that oiler works just fine, but you can also try LCM over here like they suggested. Now, one more thing to note is that this workflow has two K samplers. The first one starts at step zero and ends at the step that you specified. And then the second K sampler runs from that point onwards. But note that here, because we set it to four steps, we also need to set the end to less than four steps. So let's set this to something like two. So it stops midway. And then next it's going to use this K sampler. Now again for this second K sampler because we've linked it to self-forcing, we can also set this step count to four and then the CFG to one. And then let's also set LCM here. And then we would start at step two and end this at step four like this. And that's pretty much it. That is how you would connect this self-forcing Laura or any other Laura into your workflow. And this would allow you to generate the video in less than half the time.
By the way, one last thing to mention for all you fellow gentlemen out there. I'm sure all of you are wondering about how uncensored this is. Well, the predecessor 1 2.1 is the most uncensored video model out there. And there are luras or basically fine-tuned models which you can add on top of your workflow which can basically help you generate any action or position or fetish you can think of. Like it's really damn uncensored. Now the awesome thing is all these loras are also compatible with one 2.2. So again you can just easily plug and play all these previous uncensored loras into the workflows I showed you today.
Anyways, that sums up my review and tutorial on one 2.2. This is hands down the best free and open-source generator you can use right now. Let me know in the comments what you think of this and what other cool generations were you able to come up with. And if you run into any errors during the installation, also welcome to paste the error message in the comments below and I'll try to help you troubleshoot as much as possible. As always, I will be on the lookout for the top AI news and tools to share with you. So, if you enjoyed this video, remember to like, share, subscribe, and stay tuned for more content. Also, there's just so much happening in the world of AI every week, I can't possibly cover everything on my YouTube channel. So, to really stay up to date with all that's going on in AI, be sure to subscribe to my free weekly newsletter. The link to that will be in the description below. Thanks for watching, and I'll see you in the next one.