WeftCut
The open-source video editor your AI agent can drive
51 followers
The open-source video editor your AI agent can drive
51 followers
WeftCut is a free, open-source desktop video editor your AI agent can control through MCP. Connect your agent to trim clips, add captions, build animations from code, and edit a real timeline while you watch. Review changes, undo, or take over by hand. Available for Windows, macOS, and Linux.




Free
Launch Team / Built With

AWS Activate CreditsCredits that fuel your startup’s next milestone
Promoted


@alchemist_studio That’s close to the workflow I’m aiming for: an editor where you can cut and compose manually, or let an AI agent drive it while you watch and review.
Screen recording is planned, since I also need that for my workflow lol. But it isn’t built in yet, so it wouldn’t replace your whole setup today. The goal is to bring recording and editing together while keeping the UI simple.
@unclechair I think screen recording zoom in and out, and a good cropping tool that does rounded corners and adds edge. like cap.so is all you need. not the recording in it self. that quicktime can do. so it wouldnt be that much effort I think. I made this last week. at the bottom https://dribbble.com/shots/27742621-005 with imovie and cap.so
I tried ChatCut and wasn't impressed. Can you give me a few examples of how yours is different?
@justin_rockmore Honestly, there’s a lot of feature overlap with ChatCut. What I’m focusing on is:
• Code as video: your agent can write custom animations and graphics that stay editable on the timeline, with controls you can tweak yourself.
• Supervising the agent: watch edits happen, inspect its actions, and restore checkpoints when you don’t like the direction.
• A lightweight interface: keep the editor focused on previewing and refining, while your preferred external agent handles planning and execution.
ChatCut supports some of this too. These are our design priorities, rather than claims of exclusive features. So basically we are different from the NLE taste and some concepts.
The following GIF is an example of making an MV, the process was:
Ask agent to learn JIZURA -> make componet (named motifs in app) -> transcript the song -> use text to make it
This may not that amazing like opus 5.5 using threejs. But most of the time making a video with editable element is more valuable, we always want to changes some details if agent have some misunderstanding, and I believe the agent could help reduce those repetitive part to make creation more interesting again. :)
Hello, UncleChair ! I've never used ProductHunt before, but I've made an account just to come and tell you that I find your app absolutely amazing. Really !
A lot of different tools to make some good and easy video editing this past few months with AI models. I created something with an MCP server so that I could connect different harnesses that I use in my different jobs. I also tried Remotion, scripting, and stuff like that. But so far, I have always been disappointed. Big apps like Adobe Premiere Pro or DaVinci Resolve Studio have MCP servers who have a ton of limitations, even for simple things. Meanwhile, projects like Remotions are very interesting, but while AI can interact easily with them, it's hell to have a render that is not too long and an interface where I can easily see and make edits by myself if it's needed.
When I took a look at the website of Weftcut, I thought, "oh, here is just another of the thousands of little projects that are being made these days using AI". It honestly looked too good to be true. But as I was so in need of a software that just did the job, I tried it anyway. I've just been blown away. The interface is so clear. The options are so good, especially the fact that you can download and install local models to not only do a transcription, but also an image analysis. It's absolutely amazing !
Error messages are extremely clear, which is really different from most software I've been using; especially scripting approaches like Remotion, when everything becomes a black box that only the AI agent I'm using knows how to deal with. But honestly, the thing that surprised me the most so far is the rendering speed and the quality of the proxies. When using Remotion, I was having a lot of trouble just looking at the result because everything was extremely slow to render, even just in the interface, so that I could take a look with it. But here, everything is so fast, I'm really amazed !
Thanks a lot for creating Weftcut and for putting it into the world. I'm really impatient to see what it's going to become. I hope you will find a way to be financed while keeping the project free and open source. I think it could help a lot of video editors. At least for me, it was really a great spot between the complexity and the flexibility versus the ease of use and the great interface. Also, this feature of allowing the AI model to code, to make some animations and stuff like that, this is genius, it works so well! I've been using Cloud Cowork and it's been able to make such great animations and subtitles. I'm really extremely impressed 😄 !!
Good luck for the rest of the development. I hope this project will continue. I think you have something great on your hands and I hope you'll be able to publicize it more and so that many other people can adopt it !
I am looking for a solution for this. I work in the AI space with a youtube channel as a core part of my business so I really try all solutions for this until I find the one that really works for me. Just downloaded it.
Best of luck with the launch!
@guillermo_bascunana Thanks for trying WeftCut and for your support! We would keep improving it for better experience 🙌
@unclechair btw so far so good, cutting and trimming the raw footage at least has been better than other alternatives!