Best, legendary, interesting, funny, badass clips of your collection

[NukeFromOrbit 0]

Hi,

I'm new here on the forum and to Emby. I've been a long time Plex monthly subscriber but had to move away from Plex due to the annoying changes and direction they have chosen lately. Basically summed up by lack of focus on local media and to much "hidden" paid content/services in their UI. 

Anyway, I'm happy with Emby now. It looks good and works fine, so I'm now a monthly subscriber.

I do have a feature request which I think would make local content a lot more interesting. The idea comes from browsing YouTube. Often YouTube will recommend a clip from a move that I like. Clips like the monolog by Robin Williams in Good Will Hunting ("Your move chief") or the first encounter with Dr. Lexus in Idiocracy, or any clip from any of your movies that is interesting in some way. Either it's funny, interesting, thought provoking, scary, etc. Anything to highlight the content you have locally. It would be really nice to be able to browse your content like that. All the interesting clips from your movies, and the ability to just continue watching, start from the beginning or continue browsing clips.
This will basically address the main issue these days... "Findings something to watch"

Lately I've found myself actually looking at my local movie after being intrigued by a YouTube clip from it. So I'm thinking this would be really valuable feature to have in a media server.

I'm a developer myself so I did look into the API a bit to see if there is any obvious way to do this, but I couldn't find a good solution right away. That doesn't necessarily mean it it's not possible, but I'm thinking it's a good idea to describe the idea first here in the forum to see if this is just something I think is a good idea, or that other thinks it might be a good idea also, or if anyone have already taken a shot at it before, or if it's already implemented somewhere/somehow (or maybe the API supports something like this already but I didn't spent enough time digging into it?). 

I do see some issues that a solution would have to address. 
1) There would need to be some sort of centralized database with information, and this would also need to accept community content. So some form of reputational system for accepting input / removing spam/crap would need exist. 
2) Movies might be hosted with slightly different cut. Meaning that some might have cut parts of the opening/ending, which could make it difficult to use simple timestamps and a movie.
3) Movies comes in different versions. Like directors cut, extended, alternate endings, etc. 
 

Now 1 is a big issue. But it can be done.

2 and 3 could be compensated in two different ways that I can think of at the moment:
A)
Compensate for different versions and some logic on the movie length with common scenarios like "Lord of The Rings Two Towers", Lord of The Rings Two Towers Extended" etc. Maybe with some possibility to configure offsets if your entire collection have cut al the production logos at the beginning or credits at the end.

Include some photos/audio (or enough info in some form or another) to find the start of the clip by comparing one or more stills/audio (or something) towards the movie itself. Comparing like this would obviously not be possible with pixel by pixel comparing (using approx. timestamp and the actual movie to find the exact position), so some form of algorithm would be need. Something like this is bound to already exist and depending on the algorithm would probably make it possible to not store actual images but just some form of metadata/processed information.

So, is this something that sounds interesting? Any other issue that a noob like me have completely ignored? Could this be made? I would be happy to help.

 

Edited July 11, 2022 by NukeFromOrbit

[Cheesegeezer 3086]

It’s a mammoth task fella, myself, chef and sam are still working on ways to detect end credits and recaps for movies and tv episodes.

also throw into the mix that emby doesn’t control the content users have, like netflix or prime. They have the luxury of distributing one media file to the masses.

emby has to fight with many different versions, remuxes, re encodes of the same content with many different hashes. The database you speak of would have to be sooooo massive it could take years to populate to a usable level.

they tried this in tagchimp and chaptergrabber about 10yrs ago, to give meaningful chapters to private content  and gave up on it, funnily enough, plex hosts the old site, but doesn’t do anything with it for chaptergrabber. 

[Cheesegeezer 3086]

Sorry… btw welcome to the real world of enjoying your media lol  

[NukeFromOrbit 0]

Thanks for the reply Cheesegeezer  

I didn't know about tagchimp and chaptergrabber so that was interesting to look into, but it's also not that closely related to what I'm suggesting.
I'm a long time developer that has now moved away from day-to-day development through work, so a hobby project would be good  I've been in quite a few projects that were "impossible" and some that were "easy" and sometimes they ended up completely on the other side of the spectrum during development. So I'm very much for starting up and solving one issue at a time. Maybe it's too much, but we'll never know if it's not tried.

Thinking about your feedback it's pretty clear that the positioning would have to be assisted by some form of automated calibration. Not hashes or anything that is static. I read a bit about comparing images between different resolutions with resistance to noise and other factors. There are quite a few advanced algorithms out there, but I also saw some really simple ones that would be fun to try. Like taking a frame and resizing it to 4x4 pixels and use the color information to look for similarities. Taking a few frames within a few seconds of each other could possible be a really good "marker" for something like this. I mean, the generic timestamp would be known, so only adjustments would be needed. Something that simple might work, and it doesn't even need to be perfect, as long as it can be locally adjusted if it fails.

I'm going to try a proof of concept on that during the summer. Hopefully it works 

Lastly, I think you are wrong about it needing to be a gigantic database in order to be useful. If I start with 100 common movies then it will be useful to quite a few people. It doesn't have to be complete. It's possible to start slow and minimal and still be of use. Especially if people can add their own timestamps for clips. Just slowly grow with more movies and interesting clips.