Cool! A few of my friends have dragged me back into watching anime, so anime related side projects are fresh on my mind. This seems like a great use case for a discord bot too.
My big want is a fluff eliminator API. Like a service where people can define timestamps of where the actual good stuff is on a per-episode basis. Then a wrapper service that displays just the slimmed down episode portions chained together. But I don't know if the YouTube API or others even support stuff like that.
I suspect you could make some progress on a "fluff eliminator" by leveraging the existing "sakuga" database[1]. "Sakuga", in this context, meaning the well animated parts.
After all, there's a fairly strong correlation between high-importance scenes and overall framecount, which makes sense when you consider that frames are the main currency animation studios work with.
I don't think this approach would work. It's quite common to have long-ish dialogue scenes with lower-grade animation, and save the high-quality work on scenes with interesting things to look at happening on the screen.
As an example, the final conversation between Okabe and FB in Steins;Gate is really not good animation, but it is crucial to the plot, and the dialogue and voice acting still make it a very impactful scene.
An extreme example: There's a critical scene towards the end of Neon Genesis Evangelion where a single frame is on screen for about a minute with no dialog. (Not the elevator scene.)
You are, of course, correct. It's not exactly a bulletproof heuristic. At best, you'd probably only be able to identify likely filler episodes, as opposed to filler scenes.
A truly sophisticated approach capable of identifying filler scenes would probably involve machine learning using data that's not (to my knowledge) actually available to the public, like engagement/watchtime statistics.
> My big want is a fluff eliminator API. Like a service where people can define timestamps of where the actual good stuff is on a per-episode basis.
I'm trying to build something similiar[0] to this by indexing (and plotting) the anime score (given by MAL, something like IMDB for anime) on a daily basis. So in theory you could look at the graph and when you see a spike in score/user/favorites the episode may be good, or bad, or anything.
That's the idea. I was busy getting the indexer stable, so it's not really finished and I haven't spent much time on design. E.g. the airing dates aren't plotted yet etc. The current work is at https://anistats.com.
My big want is a fluff eliminator API. Like a service where people can define timestamps of where the actual good stuff is on a per-episode basis. Then a wrapper service that displays just the slimmed down episode portions chained together. But I don't know if the YouTube API or others even support stuff like that.