Skip Intro could use some more features

[joechilds 12]

The Skip Intro feature is great, but it could benefit from some more functionality. 

It would be great if you could disable Skip Intro for certain series. For example, Curb Your Enthusiasm has a very short intro, but has the same pieces of background music in all the episodes, so Skip Intro is often selecting one of the background songs as the intro, leading to the Skip Intro button appearing during the actual episode. 

I think there should be a setting in the Edit Metadata screen for each series, allowing you to disable Intro Skip for that series, if the feature isn’t working properly for that series.

Beyond that, the generated time codes could be saved in the NFO file for each episode, so that they could be edited manually. This would also mean the time codes are backed up. 

This would be handy for cases where Skip Intro seems to select a completely random time code as the intro. I’ve seen instances of Skip Intro completely missing the actual intro, and skipping 30 seconds of the actual episode instead. I’m assuming this is a bug. 

[ebr 14916]

 

[ebr 14916]

Just now, joechilds said:

I’m assuming this is a bug.

Not really a bug - just the fact that there is no way the process we are using to determine the intro positions can be 100% accurate in all cases.  It is a guess based on the content.  Its pretty good but not foolproof.

You may want to check out the ChapterAPI plugin as well.

[joechilds 12]

5 minutes ago, ebr said:

Not really a bug - just the fact that there is no way the process we are using to determine the intro positions can be 100% accurate in all cases.  It is a guess based on the content.  Its pretty good but not foolproof.

You may want to check out the ChapterAPI plugin as well.

Apologies, I did a search but didn’t find that particular thread, didn’t realise it was already asked for. 
 

I appreciate it can’t be accurate in all cases, but there are instances where the episode intro is identical in every episode, and Emby gets it completely wrong. For example, the intro runs from 00:30 - 00:45, and Emby marks 01:00 - 01:15 for one episode, even though the other episodes were marked correctly. 
 

The problem with the ChapterAPI plugin is there’s no way to batch-edit time codes, which you could do if they were saved to NFOs. And there’s no way to back up your changes, so you’d spend a lot of time editing the time codes, and then lose all your changes if you refreshed metadata, etc. 

Edited April 16 by joechilds

[ebr 14916]

21 hours ago, joechilds said:

but there are instances where the episode intro is identical in every episode

Identical in what way?  The process we use is based on audio, not video.

[BillOatman 500]

On 4/16/2024 at 8:54 AM, joechilds said:

I appreciate it can’t be accurate in all cases, but there are instances where the episode intro is identical in every episode, and Emby gets it completely wrong. For example, the intro runs from 00:30 - 00:45, and Emby marks 01:00 - 01:15 for one episode, even though the other episodes were marked correctly. 

I had that happen once so far as well.  I was watching a episode and the skip intro button did not pop up at the start of the intro like the others did.  Then a minute or so later, during the show content, the skip into button showed up. Only 1 episode of 1 series so far for me, but something clearly messed it up.

[joechilds 12]

3 hours ago, ebr said:

Identical in what way?  The process we use is based on audio, not video.

Identical as in the exact same audio and video. See Bill’s comment above, this is the same problem I have. 

[visproduction 123]

Doesn't Netflix and other media providers use a manual dbase entry, giving the exact number of seconds for each show?  This works for media producers, streaming online.  Trying to do the same on the fly automatically, for recorded media can probably never be as accurate.  Perhaps a setting in Metadata dashboard admin input, to make a precise number of seconds, per series, would work better.  Some episodes may have a different edit and would need a separate adjustment. But realistically, who is going to use it?   

Related forum discussion on similar issue:
https://www.reddit.com/r/PleX/comments/rvunb5/auto_skip_intro_option/

History of feature creation:
https://culturacolectiva.com/en/art/design/skip-intro-button-netflix-inspired-game-of-thrones/

 

Edited April 20 by visproduction

[bakes82 90]

And most people use pirated media of different rippers so intro detection becomes pointless for most.  Far easier to just press fwd for the 10s skip than dealing with skip intro and the multi clicks.  Chapters would work since you modify the file but you would need a plugin to read the chapters look for intro or what not and then replace the timestamps of the media item.

[BillOatman 500]

17 hours ago, bakes82 said:

And most people use pirated media of different rippers so intro detection becomes pointless for most.  Far easier to just press fwd for the 10s skip than dealing with skip intro and the multi clicks.  Chapters would work since you modify the file but you would need a plugin to read the chapters look for intro or what not and then replace the timestamps of the media item.

Far from it.  I agree a database of intros would not be of great value for the reason you stated.  But I use the intro skip all the time as it is now and it works great.  And it is one single button click that gets me to the end of the intro, not "somewhere near the end of the intro" which is what that 10 sec advance would get you. And some series would require more than one 10 second advance to even get there.

[pwhodges 1532]

The intro in anime is usually 90sec, so nine clicks (though the files often already have a suitable chapter mark, which does fine).

But skipping the ads in an ITV program takes 25 clicks on my Roku! - I wish I could change the jump to 30secs or even a minute.

Paul

[rbjtech 4275]

Introskip/ chromaprint as a technology works very well.

Yes it's not 100% accurate, but both the Plugin and to some extent the core implementation are certainly good enough.

I have mine set to auto-skip - it's that good, and it's usually within 0.5 secs of the content start before playing - that's pretty darn good in my book.

If you do have a rogue entry, then I use the Plugin tools to edit it, or use ChapterAPI for the core.

btw - an online database of intro's works very well, but not in the way you are thinking.   You don't actually hold the 'timings' - as that would obviously not work - what you hold is the 'chromaprint fingerprint' of the intro - that way, you do not need to create it each time - you just need to 'detect' the intro on your version of the file, using the online fingerprint.   It works perfectly - but it needed support from the community to upload the fingerprints - which it never got.    It's actually incorporated into the Chaper API plugin - or was.

Edited April 21 by rbjtech

[bakes82 90]

16 hours ago, rbjtech said:

You don't actually hold the 'timings' - as that would obviously not work - what you hold is the 'chromaprint fingerprint

Would be better to hold the timings much like multi versions of subtitles. You don’t need to then use cpu cycles to “find” it.  Specially useful for the numerous db issues with beta versions or large deployments.  
 

iirc in the early versions of the clients the skip intro button wasn’t forced highlighted, maybe its client dependent so you would need like 2-3 clicks just to get to the button then press it. Faster to just press skip fwd 2x and boom done. But anymore the shows I watch don’t have intros so it’s pointless.  Take 3 body problems the Netflix “intro” is like 2 seconds.  

[rbjtech 4275]

1 hour ago, bakes82 said:

Would be better to hold the timings much like multi versions of subtitles. You don’t need to then use cpu cycles to “find” it.  Specially useful for the numerous db issues with beta versions or large deployments.

Simply not feasible as you do not have any way to know which version of the file you have.  You may have the 'broadcast' version with their logo sequence inserted, you may have the 'previously on' version, you may have the DVD/BluRay version with all of this chopped out - in which case all your 'timings' on where the intro is are useless.     To find the intro vs a known fingerprint, you need to analyse the first x minutes of content.  The FP can give you some hints here to make it faster (ie rather than scan 15 minutes worth, the intro is in the first 5 minutes) , but that scan does need to happen to accurately find the intro start and end.

1 hour ago, bakes82 said:

iirc in the early versions of the clients the skip intro button wasn’t forced highlighted, maybe its client dependent so you would need like 2-3 clicks just to get to the button then press it. Faster to just press skip fwd 2x and boom done. But anymore the shows I watch don’t have intros so it’s pointless.  Take 3 body problems the Netflix “intro” is like 2 seconds.  

Intro detection has a minimum value - I believe it's 5 seconds, maybe 10 - anything less than this and it ignores it as an Intro otherwise you get too many false positives.

Edited April 22 by rbjtech

[bakes82 90]

4 hours ago, rbjtech said:

Simply not feasible as you do not have any way to know which version of the file you have.  You may have the 'broadcast' version with their logo sequence inserted, you may have the 'previously on' version, you may have the DVD/BluRay version with all of this chopped out - in which case all your 'timings' on where the intro is are useless.     To find the intro vs a known fingerprint, you need to analyse the first x minutes of content.  The FP can give you some hints here to make it faster (ie rather than scan 15 minutes worth, the intro is in the first 5 minutes) , but that scan does need to happen to accurately find the intro start and end.

Intro detection has a minimum value - I believe it's 5 seconds, maybe 10 - anything less than this and it ignores it as an Intro otherwise you get too many false positives.

I’m well aware of how it works lol.  I’m pointing out in alot of cases it’s not needed.  You can achieve the same results with the file name/hash. Just like subtitles. You search based off the file name it say this one is close or you can use any other variety there is.  
 

I believe plex stores the timing via file name, hash, media lengths, etc so if someone else does the scan and you have the same file name or what not it will pull it down, or it acts as your backup if you sync it up but you can have your system only use values from the internet and do 0 scanning.

 

 

[rbjtech 4275]

13 hours ago, bakes82 said:

I’m well aware of how it works lol.  I’m pointing out in alot of cases it’s not needed.  You can achieve the same results with the file name/hash. Just like subtitles. You search based off the file name it say this one is close or you can use any other variety there is.  
 

I believe plex stores the timing via file name, hash, media lengths, etc so if someone else does the scan and you have the same file name or what not it will pull it down, or it acts as your backup if you sync it up but you can have your system only use values from the internet and do 0 scanning.

 

 

I guess it's a viable alternative if you use the same source,  but for any self made/ripped/edited/modified file - it will obviously will not work.