I'm Alvaro, one of the founders of Mago. With my co-founder Alex (creator of Warpfusion, open source video to video), we built an open source AI video studio specialized in long form content and large volumes of video.
It's a desktop app that you can run for free on your own GPU, that uses a portable ComfyUI or can be plugged into an existing Comfy setup. It handles both hybrid production (video to video, AI VFX) and full AI (microdramas, AI series for social etc).
Back when Alex was working on Warpfusion, we were discussing making an app for "real" production, meaning hours of video, films, series, etc. We started off building a closed source solution, but we're now going back to our roots.
In short: there's a billion AI video tools and it seems like everyone wants to make their own, in fact with the progress of LLMs I'm more and more convinced by the idea of personal software. AI series for TikTok and VFX on a feature film have completely different needs.
AI video is also quite expensive still, unless you can run it on your own GPU. So not only can you run models for free, if the app is open you can also fork it into whatever you want!
Finally, we've been hearing for more than a year that Comfy can be a little too complex, so we thought, well why not try and make open source plug and play? That's the end goal.
Limitations: you'll need an Nvidia GPU with 16GB+ of VRAM and 60GB+ of disk space to host the models (Qwen, Minimax H3). License is AGPL. You can also use closed source models with a Mago subscription.
Please try it out and let me know what we should add next!
Comments URL: [https://news.ycombinator.com/item?id=50341649](https://news.ycombinator.com/item?id=50341649)
Points: 1