Enjoy.
Qwen3-VL can scan two-hour videos and pinpoint nearly every detail - https://www.hackerneue.com/item?id=46094606 - December 2025
I'm also reminded of those posters in the mall, where from a distance, it's an actor's face. But when you walk up close, it's actually made of hundreds of stills from their films.
You start with a digital copy of the movie I'm starting from...
I extract the audio...
And then my script downloads the Public Domain movies...
...and I've got a hard-coded list of cuts to take from each of the Public Domain movies...
...and combine it with the movie audio?
My script is just a few hundred KB.
As long as the movie you start from is pretty well in audio synch with the version I start from...
Heck, I could probably even compensate for that, too... A little bit of analysis to listen for the first words spoken...
https://lintangwisesa.github.io/MediaPipe-in-JavaScript/inde...
https://github.com/cosyneco/MediaPipe.NET
https://ai.google.dev/edge/mediapipe/solutions/guide
I'd think you'd start with a tool to automatically cut video into scenes, and kind of go from there...
And then to take a movie like The Matrix, or Star Wars, and then recreate it, shot for shot, as closely as I can, using clips from Public Domain movies.
Especially if I can pick, like, "this actor in this public domain movie, or movies, would be a good Neo, or Luke." And it turns out that he's in enough shots, and poses, that I'm able to remake all of that character's scenes, from Public Domain footage of this other actor.