purplemyst / caerus Goto Github PK

Right now, the only "shaving" method is removing portions of video completely from the file.

I think there's a few more methods that could be added.

• Speeding up by a given percentage.
• Adding chapters to the video file

As per the title, preserving subtitles is important.

I do not think that this is a job that ffmpeg knows how to do. We might have to find another library for ass file manipulation and the like.

As per the title, the name of the series associated with a video file is gathered utilizing filebot -mediainfo. This dependency is not one I particularly enjoy as filebot is a paid-for product.

Here's a few things to do to decouple:

• Avoid calling into filebot -mediainfo, preferring to read extended attributes directly. We can take as reference filebot's own xattr.py to see how to read them in a cross-platform way.
• Allow passing in a --series argument to manually specify the series.
• Perhaps more complex configuration such as a custom command + JSON path to get the series name?

As of the writing of this issue, the only "analyzation" method is manually marking videos.

I'd like to introduce a few more methods, such as:

• Analyzing a whole season (or potentially series) via comparing image hashes of each episode's frames. The most common frames, excluding fully black ones, will probably be openings/endings (or, more generally, fluff).
• Analyzing a whole season similarly as the previous item but with audio hashes, a la Shazam.

Detecting credits is difficult. The fact that they are mostly black makes them 90% similiar to any another frame that's mostly black, making the current searching method fail.

An interesting approach might be to special-case "black screen credits", utilizing things such as OCR to remove credits.

Introduce more mark management options:

• Define new markings a video file (and, transitively, on its series)
• Show markings currently defined on a video file in particular
• Show markings currently defined on a video file's series
• Remove markings from a video file or a series

I'd like to be able to utilize this without modifying the original video file. Furthermore, I like to utilize Jellyfin as a media player.

So here's the idea: add a mode where a given video is just analyzed and its segments are added to the database.

Then we can utilize that information in the database to simulate being a SyncPlay client, as in pretending to be another client playing the media while our user is actually playing it. We can use that to wait for our user to reach a fluff segment and fast-forward playback to after it.

I'm not sure how close to even doable this is, but it sounds interesting and doable.

Removing segments currently seems to leave over one frame or two of the original video. I'd like to avoid this.

https://www.python.org/dev/peps/pep-0585/