How to Make 2D Animation Videos With AI (No Experience)

追加: 2026-05-27

[00:00:00]It took me less than 15 minutes to create an entire 2D animation using AI.

[00:00:04]And the craziest part is that I did it all without any drawing or technical skills at all. And today, I'm going to show you the exact workflow I followed to get from a simple idea to the final result. First, we need to get anywhere between three to five strong reference images of the animation style we want.

[00:00:20]So, you can go ahead and take a screenshot of your favorite animation, show, or video that has the style you want to replicate. However, take a few good minutes to do this because the better the reference images are, the better the result the AI is going to create. Now, here are the three images I picked from one of my favorite series.

[00:00:35]So, what we need to do now is put them all together into one mood board. For this, I'll be using Higsfield, which is an all-in-one platform that gives me access to all the AI tools I need. So, if you want to follow along with me, I'll leave a link in the description down below. Once we're inside Higsfield, let's head over to the image section.

[00:00:51]Here on the bottom left, you can choose the model for the image generation. I'm going with GPT image 2 because it's one of the best out there. However, Higsfield gives you access to all the other popular models like Nano Banana Pro, Seedream, and many more. After we select a model, we'll set the quality to high so we can get the best results. And one thing I want to mention here is that this setting is different from the resolution. This quality setting is actually a new feature inside Higsfield that tells the AI model to think for longer, which will give you better results. So, that's why I'm leaving it on high. Now, I'll also set the resolution to 4K. And if you look on the left side, you'll see a plus sign. Let's click on it. And we're going to upload our three images. Now, before we hit generate, we need to write a prompt so the AI knows what we wanted to do with the images we just uploaded. So, here's exactly what I'll say. Now, let's hit generate. And here's the result we get back. The AI turned our images into one clean mood board that we can now use as the style reference for every generation that comes next. So, now the AI knows the animation style we want. However, before we can go ahead and create anything else, we first need to have a story. And I'm not talking about a vague idea of what's about to happen. I'm actually going to show you how to get a real detailed story in less than 2 minutes. But for this, we're not going to write the story ourselves. We're going to use Claude or any other AI chatbot you like with a specific prompt template I made. So, let's open a new chat inside Claude and paste it in. By the way, I'll leave this prompt in the description below. So, you can simply copy it for yourself. Now, let's take a look at what's inside. The first thing you'll see is this placeholder where you simply need to write your story concept.

[00:02:19]That's the only thing you need to modify inside this prompt. However, the AI chatbot is going to get a few more instructions. As you can see here, I told it to create a full story starting from our idea and then break it down into three different parts. First, the setup, then the action beat, and finally, a cool ending. Now, if you're wondering how I came up with this structure, it's the same one Disney uses in their most popular movies. And one more thing I asked it to do is to add visual descriptions for every part. So, the next AI we're going to be using will know exactly what's happening and how each scene should look. So, let's hit enter and see what we get. And Claude created just what we asked for. Here are three scenes. In the first one, my character is studying a blueprint outside the vault. In the second one, the alarm goes off and he starts panicking. And in the third one, he escapes and goes onto the rooftop. So, now that we have a detailed story, we can copy and paste. We can move on onto the first piece of the puzzle, which is actually creating the characters. For this, I'll head back into Higsfield inside the image section and make sure GPT image 2 is still selected. So, I'll keep all the settings the same as before, but now I'm going to click on the plus sign on the left and upload two images. The first one is the mood board we created earlier. And the second one is a photo of myself because I'm going to be the main character inside the animation. So, the cool part about this workflow is that you can actually use real people as your characters. And because we uploaded the mood board, the AI generator will turn us into an animated character that matches the style we picked. Because if I generate just one single image of the character, the results might be unstable. I want to make sure it stays completely consistent across everything. So, that's why I'm going to use a small trick. I'll tell it to generate a character sheet with three different panels. One with the front view of my character, one with the back view, and one with the face close-up. In this way, the AI will know how my character looks from every angle, so it doesn't have to guess and mess something up. So, here's the exact prompt I'll use. And now, let's click generate. As you can see, the AI turned me into a 2D character that matches our mood board exactly, and it also added all the angles I asked it for. So from now on, we can simply reference this character in any generation and the AI will make sure to draw us just like we look here.

[00:04:18]Now I want to generate two more characters. However, you need to be careful here because whenever you try to go above three characters in a scene, the AI starts to behave a little bit weird. So try to keep it under three characters total. For the second one, I don't have a real photo. So what I'll do instead is upload the mood board again and write a description of how I want the character to look. With the same settings, I'll hit generate. And here's how my second character came out. It has that nerdy tech guy feeling, which is exactly what I asked for. So, it did a great job respecting my prompt. I'll save this one as Milo. And for the third character, I want to create the security guard who comes after me when the alarm goes off. So, here's what I'll write for the prompt. And now, if we press generate, we should get a big guy with a buzzcut who looks quite angry. So, let's see how the AI handles it. Well, I really like how it came out. It matches perfectly with the animation style, and it captured the intimidating feel I wanted for this character. So, now that we have these characters in place, we can go ahead and create the next piece of the puzzle, which is the locations.

[00:05:12]For this, we are going to use a really similar process. However, this time, we don't need to get multiple angles of the same location because they'll stay pretty much the same throughout the video. So, the first one I'm going to create is the vault corridor where my character is waiting in the first scene.

[00:05:26]As I showed you previously, I'll upload the mood board as the reference by clicking the plus button, and then I'll simply paste in the prompt. Now, these prompts I'm showing you were all generated with Claude using a really similar method to the one I showed you while creating the story in the beginning, and they're really good. As you can see, they tell the AI to use the mood board as the art style reference, and they also include more details like the colors we want. So, let's go ahead and see what it's going to create for us. Well, it looks exactly like I imagined. It has that simplistic style of those 2D animations. So, I'll save it and go generate the second location, which is going to be a rooftop. Now, I need this one for the last scene where my character escapes the building. And here's what I get back after I write this prompt. Honestly, I really like how it came out. It gives me that nice mood of a city at night. So, I'll leave it as it is. Now, there's one more thing I want to create before moving on, and that's the bag that character uses throughout the entire video. You see, whenever you have assets that you use multiple times inside your video, you want to create a separate reference image for them. However, if you're using an object only once, there's no point in doing this. But for my example, this is necessary. So, just like we did with the characters, we're going to create multiple angles for it. So, the AI will know exactly how it looks from every camera angle. So, for this, we'll simply upload the mood board and use this prompt. And as you can see, Higsfield gave us a really clean sheet for the back. So, now I can save it and finally show you how to put all of this together and generate your first video animation.

[00:06:44]So, let's click the video section and choose Cance 2.0 as the AI model because this is the best video generator for this kind of animation. First, we'll upload our reference images, the mood board. my character, Yuri, Milo as the second character, and the vault location. Next is the prompt. Before I show you the one we'll be using, there are two things to keep in mind. The first one is that we don't want to create our video at 60 frames per second, which is how AI usually creates the videos. And we don't want that because it will speed up the video and make it feel too smooth for an animation. So that's why we are going to explicitly tell the video generator to use 12 frames per second so we can get that handdrawn feeling. Now, the second thing you want to pay attention to is the number of shots. You see, every animation has multiple shots, but if you don't mention this inside the prompt, the AI might try to guess what you want and end up with bad results. So, to avoid this, we are going to tell it exactly when to move on to the next shot using time stamps. So, let me show you how it actually looks for the first scene. As you can see on the top, we have the visual style. Then, we mentioned the audio and the lighting details. We also included the identity, which tells the generator which characters to include. And finally, we have the scene itself broken down into the timestamped beats I was telling you about. For the rest of the settings, I'll set the aspect ratio to 16x9 and choose the highest resolution, which is 1080p. So, let me hit generate. And here's what we get.

[00:08:03]Yuri, you there? We're live.

[00:08:14]It honestly looks really good. The intensity builds from start to finish and the multiple shots make the whole thing feel just like a regular 2D animation. But what I really like is the last few seconds where we have these slow pushin on my character's face. Now for the second scene, I'll use a similar process here. I want to add more intensity to the video. So what I'll do is set up the alarm to go off and my character will start panicking even more. So let me paste in a similar prompt and see how the scene comes out with more emotion. And here's the result.

[00:08:47]The first thing I noticed is that the location and the character look exactly the same as in the first scene. So this means our approach of building all the pieces together at the beginning was really effective. Besides that, Seedance really nailed the details. Now for the third scene, all I need to do is use a similar approach and update the references. This time I'll upload the mood board, Yuri Dex, the bag, and the vault corridor as the location reference. I won't include Milo. He's not going to appear in this scene. As you can see, this one is going to be 15 seconds across nine timestamped beats.

[00:09:17]So, it will be quite fast. And here's what we get after the generation.

[00:09:35]The motion is really smooth and my character runs just like a normal person would. And Sidance handles the emotions really well, too. First, my character is worried when he sees the bodyguard, and then he gets angry and tries to escape.

[00:09:46]Now, for the fourth and final scene, I'll change the references to just the mood board, my character, the bag, and the rooftop. I'll paste in the prompt and set it to 15 seconds.

[00:10:09]Now, what I really like about this one is that at the beginning when he goes up these stairs, the camera shakes intentionally, which makes the whole thing feel way more realistic. And once he gets on the rooftop, the location matches exactly what we generated earlier, so everything stays consistent from start to finish. And now that we have all four scenes, let's go and stitch them together into the final animation. For this, all we need to do is open Cap Cut and drag and drop all four clips onto the timeline in order.

[00:10:34]Once they're in, I'll go through each one and trim any frames at the start and end that feel off. Then I'll add the music track underneath the full timeline so it works really well with the pacing of the story. And finally, I'll export the video at 1080p. Now, let's take a look at it.

[00:10:51]Yuri, you there? We're live.

[00:11:03]A heat This result has actually come out looking super clean. So, if you want to create a similar animation without spending time learning how to draw or any hard technical skills, go sign up to Higsfield using the link in the description and create your own animation with AI right now. Thanks for watching and I'll see you in the next one.