Inspiration
Honestly, this whole project started from a place of personal frustration. I love telling stories, but every time I opened a traditional video editor, I felt like my creativity was being forced into a cage. I had all these ideas on my whiteboard, a web of interconnected thoughts, but the editor only gave me one single, rigid timeline. It felt like trying to write a novel on a single line of paper.
What it does
At its heart, InsomnIA is a new way to create videos. Instead of starting with a blank, intimidating timeline, you start on a visual canvas. You map out your story with simple blocks, and connect them to build your narrative flow. But the real magic is how you delegate the hard work. You can drag specialized AI "agents" onto your blocks. Connect the "Subtitle Agent" to generate captions. Connect the "Silence Remover" to automatically tighten up your pacing. You're not just clicking buttons; you're directing a team of AI assistants to bring your vision to life.
How we built it
This was a full-stack challenge. The frontend is built with React and TypeScript, using ReactFlow for the visual node-based canvas. It's a modern, responsive interface designed for creative flow. The backend is a Python FastAPI server that handles the initial, heavy-duty video analysis. When a user uploads a video, it uses tools like PySceneDetect and FFmpeg to break it down into logical scenes and create proxy versions for smooth editing.
Challenges we ran into
The biggest challenge, by far, was achieving the real-time, bidirectional sync between the visual Story Web and the traditional timeline. Making them two separate views of the exact same underlying data, without any lag or glitches, was a massive architectural puzzle. Early prototypes were clunky and would break easily. We had to completely rethink our state management and build a custom event bus system to make it feel seamless.
Accomplishments that we're proud of
I'm also really proud of our move to a non-destructive architecture. It was a difficult refactor, but it's the foundation for everything else. It's what makes the app fast, flexible, and ready for the future. And seeing the first local, in-browser AI agent (the Silence Remover) work without needing a server call was a huge win. It proved our vision for an accessible, free-tier-powered ecosystem was possible.
What we learned
The biggest lesson was that the workflow is more important than the features. A user will forgive a missing feature, but they will never forgive a workflow that feels frustrating or unintuitive. We learned to obsess over the "how" just as much as the "what."
What's next for InsomnIA
This is just the beginning. Our immediate next step is to launch our private beta with the "Founding Creator Circle" we're assembling. Their feedback will be critical as we polish the product for a public launch. Beyond that, our roadmap is focused on two things: Expanding the AI Agent Ecosystem: We'll be integrating more free and powerful AI services for tasks like color grading, audio enhancement, and finding B-roll footage. Building the Community: We plan to open up our platform with a developer API, allowing anyone to build and share their own AI agents. We envision a future where InsomnIA isn't just a tool, but a thriving marketplace of creative AI solutions.
Log in or sign up for Devpost to join the conversation.