Transcription
Welcome to Sandbox, the AI powered animation generator for After Effects and Premiere Pro. In this tutorial, I will show you how to install the tools and set the live server feature and walk through every single feature with practical examples. So, let's dive in.
First, you will need to install the sandbox. It comes as a CXP file. So, you need a ZXP installer like this one from aescripts.com. So, this is an installer to install the ZXP tool. Just double click on it and it will open the ZXP installer automatically.
Now, before doing anything, let's set up the live server. Just go to the sandbox tool and to the "More" and "Reveal in Explorer". Double click. You will see a folder called "liveserver" and two setup scripts. One for Windows and one for Mac users. Just double click on it. It will install all the dependencies automatically and set up the liveserver environment. And the last thing you have to do to finish the installing, you just need to get NodeJS and install it. This is required for the live server feature to work correctly. That's it.
Now, let's open Adobe After Effects. Let's go to the "Window" > "Extensions" and search for "Sandbox" and launch it. That's it. Here is the interface of Sandbox. It's pretty easy. There is nothing complex with it. So, first of all, let's make sure that the live server is actually working. Just click on the red "Live" button here. It should launch the Gemini live on a browser. As you see here, it's opened a local window on the browser. So, that means it's working good. Now, let's close it for now and focus on the tool. We will use it later.
Now, let's configure Sandbox. Just click on the "Settings" tab. First, choose your AI provider. Sandbox supports four providers: Google Gemini, OpenAI, Anthropic Cloud, and OpenRouter. OpenRouter gives you access to many models. So, for example, here I'm using Gemini 3 Flash. For me, it's the fastest and the best balance for speed and quality, to be honest. Here are the models for each AI provider. Next, you just have to enter the key here. You can get one for free from Google AI Studio. So, just click "Save Key". Your API key is stored securely in Windows Credential Manager or Mac OS Keychain, not in plain text. So, it's safe.
For the Pixels API key, this is optional but useful. With a Pixels API key, the AI can fetch real stock photos for animations. So, you can get one free at pixels.com.
Now, we just need to choose the animation engine or the rendering engine. Each has a different strength. Let's open it. As you see, we have HTML5 Canvas for 2D engines. This is universal, works everywhere. And p5.js, JS creative coding, great for organic animations. Pixi.js, high-performance WebGL for games, like like games. Yeah. For motion graphics, we have Lottie, professional timeline animations. And we have Matter.js. It's 2D physics simulations. For the 3D engines, we have Three.js for 3D graphics and PlayCanvas for advanced 3D engine. So, for most use cases, I will choose HTML5 Canvas. It works great. And I will keep it on this right now. You can choose the canvas size from here. I will make it landscape for now. And the frame rate, make it on 30, perfect for most projects. You can check "Transparent Background" if you need an alpha channel transparency. This is super useful for overlays and compositing. And last but not least, we have the "Output Folder" in the settings. So, you can save your project or animation to a custom location or automatically on a temp folder. So, I prefer to save it on a custom location to make sure everything is working good.
Now, everything is perfect. Settings are saved automatically. Let's start creating and generate some animations. Just close the tab of settings. Here we can write the prompt to generate. For example, let's say "a glowing blue orb that bounces around the canvas with a trail." That's it. Let's try this one and click on "Generate".
Now, it's generating the animation with Google Gemini. Mostly, it will take like 15 to 20 seconds to complete the generation. Here we have the animation appears below here. These are dynamic parameters the AI created dynamically. So, we have the speed, or the orb size, and even the glow intensity, or any control you can just say in the prompt. For example, I want a speed or a hue slider for changing the color. You can ask the AI for any slider. It will automatically generate this parameter for you as a slider. So, this is a game changer, to be honest. If you like this animation, for example, you can just click on the "Save" button and name it. Now, the animation will save permanently in your animation library with a thumbnail and a load button to load it again.
Sandbox includes professional style presets. Let's try one. Just go to the "Animation Style" here. Click on it. It will show you some presets like kinetic typography, logo reveal, or particle effects, or even audio visualizer, retro vibes. Let's try, for example, a kinetic typography. And let's add this prompt: "Create kinetic typography animation of sandbox ball dynamic and motion blur." Let's go "Generate".
Perfect. The text is animated and has a professional motion blur effect with a lot of sliders like font size, speed, intensity of the wave animation, and motion blur. Here are the strengths of motion blur. I think, yeah, you can change the word here, even. Or let's say, for example, "the future". So, yeah, that's it. You can try any preset you like and whatever you like.
Have a big idea but want to make it better? Use the "Enhance" button. This will enhance your prompt instantly. Just click on it. It takes like one second to generate an enhanced prompt of your main prompt. You can change and edit and tweak the prompt as you like. For example, you can change the color code here and whatever you like.
Okay, now let's take a look at these buttons here. We have eight buttons. Each one with a different functionality, and they are super useful. So, let's start with the "Reference". You can upload any image reference or even a video if you like. For example, I will upload a UI design panel. I have this image of a UI design. I will tell the AI to regenerate and animate this exact same design. Just upload it and say, for example, "Design and animate this exact UI panels from this uploaded reference. Make it fully interactive with my mouse." Let's click "Generate" and watch the magic.
And boom. In less than 30 seconds, you have regenerated and animated your exact same reference image. And it's fully interactive, dynamic, and has a parallax effect. As you see, when I move my mouse, you can click here on the "Full Preview" to make it with a full-resolution preview. Just click on it. Now you can see the whole animation. So, let's make it full panel. Wow, it's super smooth and fully interactive. Let's close it and go back to the main panel. If you like the animation, just click on "Apply to Timeline". It will insert it automatically. Just click on it. That's it. It's 5-second duration. You can make it longer if you like or shorter. So, let's make it for five as a default for now. Okay.
Now, let's remove the reference and talk about the "Data" button. Just click on it. It supports formats like JSON and CSV, and even plain text. For example, I have, let's say, this car CSV. But before we do anything, let's take a look at the main CSV. Here's the data. As you see, this is all the information about the car models. So, this data will automatically be sent to the AI as a reference, and based on them, it will create your animation with the best possible results. So, I will say, for example, "Create an animated bar chart showing the data of these cars with smooth transitions." Click "Generate".
So, the AI automatically parses the CSV and detects columns and creates the charts. Wow. Created with a background, as you see. Can turn it off or turn it on. Let's make it off right now, or maybe on, and click on "Full Screen". Okay. It's a background, a real background. And here's the real data from the CSV. And by the way, if you change the data here from your Excel or from a text file, it will automatically affect the animation. It's like data-driven animation. So, this is a huge step to create stunning data animation.
We have the "SVG" button. This supports SVG. As you see, I have this vector emoji. So, I will upload it. Just double click. You can add any prompt you like. For example, "Animate this emoji" or "Make the eyes blinking" or even click on this small pawn to open a new panel or a new dialogue for labeling character parts. Here we can click on specific parts of the body or even the whole design to assign labels. So, this will direct the AI to create a correct animation without any hallucinations and a correct ordering. So, just click on the head here. You can see all the parts that we have. For example, the head. That's it. It's assigned here, as you see, "head". And for the right eyebrow, just click on it. "Right eyebrow" and "left eyebrow". Even the eye. "Right eye" and "left eye". Hold Alt and click to assign anything or any name you like. For example, "pupil". Maybe "right pupil". Enter. And this is Alt and click too. Let's make it "left pupil". Enter. So, this is the mouth. Click. Let's search for the mouth and click on the tongue. Great. Now, click "Done". That's it. Let's create a very simple prompt to test our rigging animation. "Rig this emoji face." That's it. Just four words. And click "Generate".
So, here it is. As you see, the AI generated a full animation of this emoji face. Like, even the eyes are tracking the mouse or the cursor movements. If you want to change or edit anything, just type your edit here and press on this "Edit" button. It will send the same exact code and animation and then makes just the edit or the requested new things. So, with the "Label Parts" option, there is no more guessing for the AI. So, it should do the correct rigging most of the time. Let's remove the SVG.
For the "3D Model", you can upload any 3D model, for example, a car, and choose a Three.js engine to make this feature working with it. For example, I can upload a glb or glTF file, but I prefer glb. It's much better. So, let's click on it. I have this model of a BMW car. Double click. Now it's uploaded. Let's say "car racing game. Use the uploaded car model." "Generate".
Now, it should create a game, a playable game of this car model. So, here it is. As you see, you have all the sliders here, like the transforms of this car, for example, X, Y, and Z position. And even for rotation, let's change the Y rotation and make it toward the correct direction. Let's make the car bigger. Here you can change the camera angle. As you see, you can make it full screen. So, this is a real 3D model, a glb model. It's used inside Adobe After Effects directly. You can create a simple game, play it directly from After Effects, and import it as a PNG sequence footage because my tool imports all these animations or graphics as a PNG sequence. This is how Sandbox works. If I click on "Apply to Timeline", as you see, it's a PNG sequence layer. If you change anything, for example, the angle, you just want to click on "Re-record" and that's it. It will record it from this angle. "Apply to Timeline". That's it. Okay. This is for the 3D model button. Let's remove the car model and switch back to HTML5.
For the "Font", you can upload any font you need to let AI use this exact font for your animation or typography. And the "After Effects Shapes" simply takes the data from any shape you are selected from the timeline and pushes or passes this data to the AI. For example, I have this shape path. Just select it and go click on "Shapes". As you see, captured one shape with one path. Let's say, for example, "Particles that follows this shape path with full controls" and hit "Generate". Just wait a few seconds and you will see the result. These particles will follow this exact path.
Here it is. As you see, we have all these sliders like "Particle Count". This is how many particles are in this path. "Particle Size". Wow. Now it's like fire or smoke. "Flow Speed". "Trail Intensity". "Glow". It's even glowing. Can change the color hue. Just click on "Re-record". Then "Apply to Timeline". That's it. It's in the same exact position of the main path. Sometimes, sometimes it's shifted a little bit. You just need to match it a little bit. That's it. So, this is a great feature to interact with After Effects shape layers. So, yeah, that's it. This is all for these six buttons.
Now, let's talk about the "TTS". This is special for text-to-speech generator. You can generate a voiceover directly from your After Effects panel. Just click on it, write anything you want. For example, let's say "Welcome to Sandbox, the AI powered animation generator for After Effects and Premiere Pro." You can select the voice here, any voice actor. For example, this one. And here you can write or specify the style notes. Let's make it "deep and warm" or maybe "deep" or "tutorial style". "Generate Audio".
>> Welcome to Sandbox, the AI powered animation generator for After Effects and Premiere Pro. >>
Let's change to another one, this female, and click "Generate".
Welcome to Sandbox, the AI powered animation generator for After Effects and Premiere Pro.
If you like it, just click on "Download". It's automatically inserted to the timeline. So, yeah, this is the text-to-speech all-in-one. All the AI possibilities in one single place.
Before we jump to the live server feature, let's talk about other features inside the UI panel. Here you have the "Generation History". Everything you just did, you'll see it here. But after restarting, it will reset automatically. For the "Capture" and "Mouse Motion", these are very functional and useful features to interact with Adobe After Effects null layers or any animated selected layer like controllers. For example, I have this null layer. Let's do a simple animation for this null layer. Created this motion path. Let's write, for example, "Particles reacting to my mouse" and click "Generate". Nice. Here I have these particles reacting to my mouse, as you see. But the fun part is that you can select this null layer and click on "Mouse Motion". After 1 second, it will track or capture this motion as a mouse movement. In this way, Sandbox reads the keyframes and starts simulating mouse movement easily. Let's stop the motion. Now, for recording this motion path, just hold and click on this "Capture" button. Alt then click, that's it. It will restart the animation from zero and then after 1 second, based on this keyframes time, it will start recording to match the exact motion of this null layer. Just "Apply to Timeline". Let's see. It's perfect and accurate. Let's attach something here. So, yeah, that's it.
But if you want to do the reverse method, just hold Shift and click on "Capture". It will convert your mouse movement into a null layer, like the invert method. Delete everything for now. Let's hold Shift and click "Capture" and start reacting with my mouse. This mouse movement will convert into a null layer. A new purple button appeared in the bottom. It's "Export Motion to After Effects". Click on it. This is the null layer. And click "Apply for the PNG Sequence". That's it. As you see, perfect and accurate.
Or you can even add markers to any selected layer. For example, a null layer. And let's add two markers here. So, this will convert or become click events. So, for example, let's say "a radio wave animation that explodes when mouse clicking". Click "Generate". Let's delete this one. Now I have this animation of a radio wave and it explodes when I when I mouse click. But if I select this layer with these markers and click on "Mouse Motion", now it will track the same path motion. And in addition to that, it will take these markers' times as click events. Click on "Mouse Motion", as you see. Perfect. Let's stop motion and hold Alt click "Capture". Let's see how accurate is this. "Apply". Perfect. This is the first explosion at the same exact time of the marker. Here is the second one. That's it.
We have covered all things related to Sandbox except one last feature. It's the "Live Server". I think this is a really game-changer for real-time interaction with After Effects and Adobe Premiere Pro. Let's launch it. It will open this browser window automatically as a localhost server. We can connect this to Gemini. It's free 100%. Just go to the settings and paste the key here and save the configuration. That's it. Make it Gemini 2.5. It's free and super fast for now. "Save Configuration". Then just click on "Connect Gemini". That's it. It's connected to Gemini right now. We can even start sharing our screen to let the AI see what you are doing. "Share". Let's just click on "Link Python" to link this server to After Effects directly. As you see, "Connected to After Effects". Now we can enable the AI voice and talk to Gemini.
"Hey Gemini, do you hear me?"
>> Yes, I hear you. How can I help you today?
>> "Uh, what do you see right now on the screen?"
>> I see a radar graphic in After Effects with some cyan circles and smaller points of light moving around. There's also a toolbar on the right with various controls.
>> "Okay. Uh, can you delete all the layers in the timeline, please?"
>> Great. Can you create a new composition 1080 by 1920?
>> "Oh, great. Uh, create a new red solid layer. Create a text layer saying 'hello world' and animate it from left to right. Two-second duration. Apply a wiggle expression on the position of this selected layer."
>> Nice. Finally, can you create a script with a UI panel for applying a wiggle expression with some options?
>> I have executed the script to create a wiggle expression panel in After Effects. You should see a panel with sliders and an apply button.
>> "Okay, thanks for that. See you."
>> You're welcome.
So, we used Gemini as a director, an assistant to help you with everything you need inside Adobe After Effects using the ExtendScript API. By the way, if you go to the settings, you can upload skills to make the generation or the execution much better and smarter. So, I will upload a skill file for you for free and make your AI much smarter in Adobe After Effects functionalities and all methods and scripting to give you the best possible results, even with all the models with Gemini 2.5. So, yeah, this is Sandbox, and this is the future. Thanks for watching.