Today we're putting LovArt, the first AI agent created specifically for design to the test. So I just received an early access invite and I want to test it out with you to see if this thing actually works. Unlike typical type and generate tools, LovArt thinks, plans and executes by coordinating multiple AI models like an on-code design team that never needs a coffee break. You give it minimal input and it breaks the task down into steps from style strategy to final output. In this video, I'll show you exactly how it works, run a few real world tests and highlight use cases that could save you hours in your daily design flow. Let's jump right in. Here we are at the LovArt homepage and if you haven't got the invite code, you can sign up and request one invite code for yourself. On the homepage, we can see the use case like the bag design, product design, or maybe some of the brand identity design and you can also create art or poster or even comic book meme set as this is and because love art has all these tools available for it to use once it has a design task it can think through it and deploy all the available tools for it and complete your design goal as needed using love art is easy you just have to talk to it like chatgpt and you can even use a reference image or multiple reference image. You can even ask you to fine tune your design using natural language. Let's sign into the LovArt dashboard. And once you sign in, you can see there is a very clean and simple design and there's a chat bot and I think everyone knows what we can do with this. You can also attach your file and use the internet connection to search online. On the left hand side, we can see there's project where we can organize our design on the project. And under here, we can see other people's creation. These are some really cool design. If you're interested in any of the design, you can click in and then you will see the replay of how this design is generated. And you'll see the process of how it's seen through all the necessary decision points and the tools it use. and the process that it takes to generate the final output and we can see the output on the infinite canvas on the left. If you like this type of content, make sure to subscribe for more videos like this. Now let's see what LovArt can do with more complex design tasks. For the first example, I'll ask it to create game posters with this detail prompt and you'll see there will be a new tab open and LovArt will think through the goal and started to divide and conquer. You can see it's thinking process and you can talk to it on the chat interface on the right. We'll see the design result on the left. For each asset, it will give you an estimated time for creating that asset. For example, this one is two minutes. We don't have to wait, we can come back for it. So I went and grabbed a coffee and I came back and just about 10 minutes, it generated the different characters and the different styles that we specified. These are some really high quality design works. For each image, we can left click to select and further edit it. We can select a part or use the brush tool or ⁓ edit it by talking to an AI agent directly here. We can even add it to chat and ask it to further edit this. When we click the image on up top, we can see that we can upscale the image. We can outpaint it, which means to expand the border beyond what it already has. And we can remove or erase certain things from the image. We can also inpaint by assigning a certain region and ask it to change it into something else. We can also smudge by making the background blur or something you don't want to show as clearly. Let's try this is interesting. You can segment subject. you can segment subject. You can add to what is already there. Mask feather, invert area where it's not masked. And of course we can undo redo. There are all type of editing stuff you can do here. Try to use the remover to remove Mia. Remove successfully. Okay. You can see the past history from here. we can also see the workflow where I have this image and then the next step I remove it. Let's exit the workflow. And also we have this layer panel where you can ⁓ hide your layer. And when you move things around something like this, you can also put it underneath or above all the good stuff designers are probably very familiar with. And when we right click, of course, There are more things we can control like download or export. Let's see what we can export as and we can also export at different image size and format. These are pretty cool. Let's try to smudge. Adjust the brush size. Brush size. I smudge the lightning. I don't think this is a good design. Okay, restore. When you click into the image on the chat, you can actually also flip things around, change direction of your canvas, things like that. They're just... a lot of control you can do other than just telling the AI to do the next step. These are pretty fun. What makes LovArt different from the ordinary image generation tools is that it can not only conceive designs based on prompts, but also automatically call various tools and models to complete professional design generation. Combined with the canvas and editing functions, it can also intelligently fine tune the AI generated designs that it created. Now let's do something more interesting. Traditional game assets generation tools require manually selecting styles and entering prompts for each image. Let's try this prompt. ⁓ style with the theme of characters in the Lord of the Rings 22 tarot cards are generated and then I listed all the cards I want to generate and then it will just start working on the 22 assets that you're trying to generate. These are some pretty long process that we have here, but I'll get some Starbucks and then We'll come back for it. I didn't really go to Starbucks, but I did do some chores around the house. And when I came back, it's done. I didn't time how long it is, but I actually saw that there were some errors in the process when it's trying to generate images. There's some content policy restrictions, but it actually tried to fix things on its end. And eventually it actually complete all 22 cards. Even with all the errors. and obstacles in the middle. It actually solves problems for you. If you ever generate images with ChatGPT you know there will be some errors in the middle. But luckily you don't have to deal with it since your AI agent can do it for you now. And these are really, really cool images. And you can see that the characters are actually looking like the... characters in the movies, LoVar automatically analyze the aspect ratio and pixel required according to the prompt and then use the model to generate a complete set of 22 images. For game designer, this is a game changer. You can quickly complete designs of game characters, props and scenes in a fraction of time it would normally take. Imagine completing an entire deck of cards when you're doing laundry or buying coffee, which I did. Lovart also let us upload reference images and batch process designs to match the style. For example, I uploaded these flat icons and asked Lovart to turn them into 3D renders. So actually the first time it didn't see that the image contains nine icons, but after I told it, it reanalyzed the image and actually complete the task perfectly. Sometimes you have to communicate with it still like... when you are talking to a designer. Although it also used GPT-40 model, LovArt makes the experience far more streamlined than chat GPT, which usually creates only one or two images at a time. It understood visual context and produces smart variation with almost no input needed. To help with design and post editing, LovArt supports generating images with multiple layers. Just add editable text layers in your prompt. So what it did is it create an image of the poster and then it turned this image into something with editable text layer like this. The text is a little bit messed up when it is extracted. There's some way for it to go. I'm sure they can easily improve this and maybe create something like canvas where you can actually extract things in the right place instead of having the text running around like this. But the poster is pretty cool here. For the text, you can see all the extracted text in the layer and you can further edit it. If your product image does not have as much wording, I think it will work. LovArt also integrates the clean video model supporting conversion of generated images into videos. Let me show you an example. First, I'll ask the agent to generate Lo-Fi music covers. It created the cover images for us. Let's try to see if it can turn the images into videos. So it says that it will use clean video generation and it is using clean 1.6. Not sure if we can choose the model. It actually generated the videos. Let's see a few. These are pretty good for lo-fi or cover. This one is pretty good for lo-fi cover. We can loop the five second clip into longer ones. I think this one is pretty cool. A little bit wonky on the leg side, but this is cool. Nothing is really moving in here. Is this one moving? We used only two prompts and it generated five videos for us that we can use directly on our cover. Well, except for the last one because it didn't move and that happened. To easily download the file, can click the file icon and then you can download your files here. We can also easily share this conversation with other people by copying this link and other people will see the whole conversation replay in front of them. We can also create consistent character stories and even turn it into a movie. Like I just give it a very simple prompt here about a five scene storyboard of a boy who dreams of flying. And I asked it to keep the boy consistent and then turn it into a short movie with voiceover and we'll see what happened. Okay, this one is crazy. a little bit too much like so so at first think about the story the style and what would be the character look like what are the other characters in the story and then it create the scene just like what we will do if we're doing it ourselves or if we're asking ChatGPT to do it for us but then you can see that it gives a lot more details. It evaluate the requirement and it looks for the LoRa that it will use for the story. It did go through all the different styles and then it started to create a storyboard. We can see it created three characters and it created five images with Flux. This step, I wasn't very sure why it did that. It created a HTML file about... the planning process. So it's easier for us to see what's the planning. You can see there's characters and these are interactive. It's probably very easy for AI to do this, but as a human, I'm amazed. It planned shot one, how many seconds there should be, zoom in, zoom out, narration and some highlight. So it actually created a website to plan the whole thing. I don't know if it's for me. I feel so touched. And they proceeded to use clean to animate these images. And then it created a background music, Oliver's flying dream. This is very cool. Let's take a listen. You ⁓ And then it creates the voiceover narration. And then it uses the video editor to actually stitch everything together with the background music and also the narration. It also gave me a long report of what it did and what kind of quality control it did. And the task is completed. I didn't have to put any more input other than first... prompt that I gave it. Okay, let's watch this together. I do think that a lot of the scene can be improved but I'm very surprised how the whole process really didn't take me anything other than the prompt and there seems to be some lip syncing here but I wonder if that is lip sync because it's not really syncing and just mouse movement. Probably it's not, it's just animation. This is definitely something that will improve as time goes and it will be so much easier in the future. We can have less interaction with it and it will have better quality control itself. And this is the future guys. We should probably all consider adopting a AI agent. So there you have it. I'm shocked every day about the daily AI advancement. And this is definitely one of the shock. If you want to try LovArt, you can click the link in the description below to apply for access and drop a comment to let me know which feature of LovArt impressed you the most. Thanks for watching and I'll see you in the next video.