Description A video editor built around tight human-AI collaboration — the user provides creative direction, the AI does the boring mechanical work (arranging, cutting, syncing). Built to be accessible and open: free to use, bring your own model. Inspired by how Cursor works for code, but for video. Not input→AI→output like every other AI editor. A real timeline with a real playhead, where AI suggestions are edits you approve, not a finished product handed to you.
AI Video Editor I am really struggling right now with the implementation. I have tried 3 different versions now, and they all end up hitting a wall because the underlying architecture doesn't support it. I just need to narrow the scope more and more until the core functionality works.
AI Video Editor After researching real editor workflows and pain points, I’m refocusing the project on the core problem editing long, multi-track content quickly, without sacrificing control
Two north starts for this direction:
Save the user as much time as possible. The main pain for my target users (like content org esports editors) is the amount of time spent on syncing, sorting, and rough cutting hours of footage.
Give the user full control. Most AI editors take the tracks as input, and give you a finished output. My bet is the opposite: AI that does the repetitive, deterministic, boring work like syncing tracks, switching between tracks, and lets the user focus on the creative work like zooming, music, text, images.
This means the MVP will focus on fast, rough editing of long multi-track content, with clean exports to more detailed editing tools like premier pro. The goal is to solve the biggest workflow bottlenecks for editors who care about both speed and control.
AI Video Editor Big update
I have done 10 detailed expiris, and proved that:
- There is a real business case for this, the cost is lower than I thought.
- Lossless cutting fits this use case perfectly, and works in browser and an electron app.
- Tools like transcription, motion detection, synchronization work quite well for giving the AI enough context.
AI Video Editor Huge progress!
I had to switch to a real video editor, and found Olive was a great base. It has a GPL-3 license, meaning I can monetize and make a product out of this. The catch is that I have to have the source code of the editor public, but not the server, which is completely fine!
Timeline
I even have a way to export the timeline as well! It's insane how good cursor is.
AI Video Editor Expiri
It's absolutely insane how fast I got a version up and running. This already feels better than all other AI editing softwares on the market, because it gives you control to tweak, approve and reject changes. I can make manual edits however I want, but most of the bulk is done by the AI.
I also have a new insight from listening to Lenny's podcast that I am implementing here:
Behind the scenes, the AI product should get more complex, better, handle more and more responsibilities. But the front-facing interface should stay the complete same for the user.
Also, the user should still feel like they're the ones editing, they should feel in control of what happens. There are so many annoying AI video editors out there that just give you a finished product, and it's never how you want it. AI should be used to amplify the user, not take their job.
Here's a screenshot of the first version:
