2024-10-28, 01:02 PM
(2024-10-17, 06:36 PM)TheDreadPirate Wrote: The way the intro skipper plugin worked is by analyzing and "fingerprinting" the video looking for where there are repeats between episodes (intros/outros). But that is computationally expensive, in additional to all the other features in Jellyfin that are computationally expensive. In 10.10, there will be a feature called "media segments" that may enable a built-in intro skipper at a future time in a much less computationally expensive manner.
Seems you may unintentionally be spreading a bit of misinformation, since Intro Skipper is one of the providers for the 10.10 API integration. Jellyfin has replaced the need for us to inject a button by providing an endpoint to design native buttons around. The process itself, though, is fundamentally identical with a few added steps.
1. Jellyfin’s method is just as, if not more so, computationally expensive because it separated collecting and recording the segments into two isolated processes that require sequential writing of data.
2. The method of analysis you claim Intro Skipper uses is actually trademarked by Netflix. We analyze the audio, which is significantly less expensive than analyzing the video. The various analysis methods are explained in detail on our wiki (https://github.com/intro-skipper/intro-s...tion-types)
3. Jellyfin’s chapter segment provider is an identical process to our first analysis pass, so any compatible media is not subjected to further analysis. However, unlike Jellyfin’s chapter segment provider, the other videos (80% on average) that go undetected have a chance.
If you have questions about how Intro Skipper works or need more clarification and can’t find it on our GitHub repository, please feel free to ask.