Google's VEO: The Cheapest Way to Use the Best AI Video

Recently I have started looking at AI video generators for some of my personal projects and have found VO2 from Google to be one of the best out there. However, it's extremely expensive. Even on their own platforms. The price ranges from 35 all the way up to $0.50, which makes it extremely expensive for any bigger projects. However, recently LTX Studio integrated VO2 into their platform and they're offering it at a much reasonable price of almost 8 cents per second. Okay, so here's the platform and it's very interesting, especially for people who are interested in AI video and image generation. And they are also the sponsor of today's video. Now, apart from image and video generation, they also have this very neat feature of AI powered videos. So, so you can create short films or commercials using this tool. All you need to do is just describe what your story is about. You need to just provide some details about the format, the tone, genre. The AI system will help you create a storyline and then based on that it will cast characters. You can bring your own characters if you want. So here's something that I previously created. It has every scene along with static shot or frame. The character consistency is pretty good. Now if you want to animate a specific scene, just click Create on Generate motion. So here is, for example, I created a video based on an image the system generated, right? And it is pretty good. Now all of this is powered by multiple AI tools, but there are two which plays critical role. One is the image generation capability and the second one is video generation. And you can try some of the models for free on the platform. So there are two options. One is Generate images and the other one is Generate Motion. For motion generation they have the LTX video model, which is an open source model. I covered that in a previous video. And they also have VO2 which is very reasonably priced compared to anything else out there in the rest of the video. I'm going to show you my own workflow of video creation. I usually use a reference image for my videos because that gives you a lot of control. So we're going to go from a reference image to an AI generated video. As you can see, I have been playing around with this model and I think it does a really good job of creating realistic images. Even it can create images of celebrities. Some of the other models don't allow you to do and the images are pretty realistic. It can also render text for you. So that's a pretty interesting capability to have. Okay, so we're going to start with this prompt. It's a highly detailed Prompt. But essentially we want a photograph of Joker and we want an explosion in the background. So think of this as a scene out of a movie. Okay, so generated six different images. And I can actually see. See, it's sticking to the prompt. The quality of all of these images seems to be pretty good. Also, you can upscale the images if you want, but they have this neat feature of predefined styles which gives you a lot of control. So let me actually select one. Let's say we want the Joker image to be colored. Let's see what it comes up with. Okay, this is pretty neat. I actually really like this one. Okay, so I have been playing around with this. So here's an image of a scientist. Here's the same image in the style of an anime. Now, something I noticed with the image generation, if the human faces are away from camera and they're not focused, then sometimes you see some distortions. So for example here as well. But you can fix this with the simple upscale option. So I think I already upscaled this image. But let me show you one more example. Let's pick this image or maybe something like this because it's just a side view. So we're going to upscale this image. And now you can see that it does a really good job at faces. Right? So if you generate an image and you see some deformations, upscaling might help. Okay, but there is one more feature that I want to highlight. So let's say you select an image and then just click on Add Motion. So this will take you back to the video generation model. Here you have three different options, including VO2. So let me select VO2. You can either provide a text prompt or leave it blank. In that case, the model will itself try to generate some sort of motion in this image. Okay, so here's what it did with this image. Now, I think this is a hard one because there is fire in the background and you can actually see this. It's not a normal Joker image, but it still added this motion that this Joker is walking towards the camera. Okay, so Here is a AI video generated by VO2 based on a more realistic image of Joker. And it looks pretty fantastic. The movements are more natural. Also, the fire looks a lot more natural. And actually, I don't know if you can see, but there's even a shadow here as the Joker is walking. Now here are four other videos that were created by LTX Video, which is a 13 billion model, which is definitely much smaller compared to VO2. But also this one does a pretty decent job. Now, one of the main use case of these models is something like this. So you can create commercials especially if you provide a reference image. And I'm always fascinated by the ability of these models to create movements or motions just by providing a reference image. We're not even providing a text prompt in this case. Now here are a couple of videos with text prompt. This is generated by VO2. Here's another shot of a purse for reference. I think these are videos that I generated using LTX video. So VO2 definitely looks much better, but it's not perfect. So for example, if the prompt is too complex, sometimes it has trouble if there are way too many things in a provided image, right? So play around with it. I think it's an excellent tool that you want to use in your workflows. But it's also very important to keep the capabilities and limitations of these models. Here are some other example outputs from VO2. The quality of output is going to be dependent on the quality of your prompt. So do let me know if you are interested. I can create a video tutorial on how to prompt these video models. Some interesting capabilities of these video generation models. So in this case, even if the camera angle changes, it's able to preserve the face of the person for these two prompts. I did not provide a reference image, but still the output character is consistent, which is pretty incredible. And it's extremely hard to do in a video generation model. I still feel like video generation is a lot more difficult compared to the LLMs that we are using. VO2 is definitely one of the best AI video model out there and I'll highly recommend to check it out. VO2 on LTX is probably one of the most affordable option out there. So do check out LTX Studio and let me know what you are able to create. Anyways, I hope you found this video useful. Thanks for watching and as always, see you in the next one.