I Built an AI Video Generator in 10 Minutes (No Coding Required!)

I built a video generator using AI in just 10 minutes. Here's how. For context, this is using ReMotion, a framework for building video using React. I was so excited the first time I saw a demo of this. It sounded great in theory, but in practice, I just couldn't figure out the physics, the timing, and how to translate what was in my head to the code that it needed to work. Honestly, I spent an embarrassing amount of time just trying to get basic animations done. But I figured out a better way. I connected OpenCode to the ReMotion MCP server, and I was off to the races in minutes. Let's take a look. With the ReMotion MCP server, it gives your agents full access to the ReMotion documentation. So it's not guessing at APIs, it's looking them up in real time. With the MCP server set up with just a couple of lines of JSON, I can start to describe the video that I want. Dark background, orange glow, animated headline, a code snippet, and then a text underneath that. Now from there, the AI agent was able to scaffold the project, create the compositions, and configure the spring animations. All the stuff I could never get right manually. And here's what it looks like. And just like that, we have a working video to get started with. Now this is the part where I would typically spend hours trying to get the next level of detail correct, but this is where we can iterate really quickly using our AI agent. A first draft is never good enough, so here's where this really shines. Move the logo to the bottom left. Done. Make the animation snappier. Done. Add a dramatic entrance on the last headline word, scale it in with blur and a glow. Done. It adjusted the spring stiffness and damping. I don't fully know what that is, but I don't really have to. I just care that it worked. Each change takes seconds to iterate on. I don't have to look up the documentation. I just have to speak in plain English. But here's the one that really got me. Start zoomed in on the code snippet and then zoom out to reveal everything else. That's a relatively complex thing to understand. Multiple timing dependencies, but the AI handled it in one shot. So now the gap between I know what I want and what I actually see on screen disappears. But we can take this even further. We've got one great video, but what if we wanted 10 more? So here's my next prompt. Refactor this so the text content is configurable. I want to create new videos by swapping out the headline, code snippet, and subtitle. So now the entire video is driven by one object. Headline words, code snippet, subtitle. That's it. One additional thing we might want is two different versions of each video. One for desktop and one for mobile. So horizontal and vertical. So let's give this another prompt to give us a vertical template as well so that we can reuse these for videos that we create in the future. And here's what it looks like. With that in place, let's build announcement videos for the registrar API. Title is search, verify, buy. And the snippet shows a path to the API endpoint for the registrar API. Boom, it's ready. AI updates the props, renders it, landscape, and vertical. Now the last example of this is that I had my agent go and parse all the announcements from agents week last week. And then I told it to take each of those and convert it to the text necessary for a video. With that, I was able to generate all of these individual videos in just a couple of seconds. You probably noticed that there were a ton of announcements during agents week. So I'll have a link in the description below so you can find out more details for each one. So 10 minutes ago, I had nothing. But now I have a repeatable and reusable video generator. And I never had to understand any of the physics behind the scene. The ReMotion MCP server gave AI the context for the documentation that it needed. And the rest is history. So let's take a look at how this works. Can you see any other fun use cases for using the MCP server with ReMotion? Let me know in the comments below.