Completed Show Intro Skip Option

[Cheesegeezer 3086]

4 minutes ago, rbjtech said:

Thanks - yes the design is to always ensure the Intro Start is a little 'late' rather than early and subsequently the Intro End is a little 'early' - then if you have the Auto Skip enabled (when it comes..) then you don't ever miss real content. 

On recaps - yes, this is tough one as they are impossible to automatically detect.  The recap (if it exists) will almost certainly start at 0:00 and if the Intro immediately follows - then it's fairly easy to deduce the 'recap' part.  However - a lot of shows have the Intro partially way though the show - so now we have no way of knowing what is a recap and what is 'new' content before the Intro starts.

It also of course also depends on the source - Premier TV recordings generally have a recap, but repeats/DVD/Blu-Ray versions will not.

During the development, we had an 'include recap' option per show (default to off) and if your show Intro was directly after the recap - then by turning this on, you skipped the recaps too but it caused possible confusion for the initial release - so it was disabled.

Once we have completed the main Intro and End credit functions - then this can possibly be added back in -  @Cheesegeezer/ @chef ?

 

Jeez, i'm like a parrot.  We posted at the same time.  But at least we are on the same page.... hahaha  

Probably a good think considering we are the dev team for this haha

[fallenwitch3r 41]

I run a low powered machine with a Intel j1900 processor. My scan takes around 30 hours with 77 Series and around 2000 episodes. The intro and the end credits scenes are good identified. I would really love to see the skip GUI feature.

Is the GUI feature version multilingual?

[Cheesegeezer 3086]

55 minutes ago, fallenwitch3r said:

I run a low powered machine with a Intel j1900 processor. My scan takes around 30 hours with 77 Series and around 2000 episodes. The intro and the end credits scenes are good identified. I would really love to see the skip GUI feature.

Is the GUI feature version multilingual?

I think we should just explain what's happening here.

You are currently part of beta testing this plugin.  This is not released yet on the server plugins.

The GUI feature as you call it, will be implemented by Emby @Admin as this requires core features to be updated which we don't have access to.

We really do appreciate your testing and feedback and the more successful scans we get the quicker this will be implemented.

Once your scans are complete can you run the Statistics task and then check the stats page please or upload the stats.txt file to here.

[chef 3745]

1 hour ago, fallenwitch3r said:

The accuracy is really good. Sometimes there is a delay between 0,5 - 1,0 Second after the intro starts. Skipping reacaps  could be a nice feature..

Im interested if you got your results with fast detect on or off? 

Thanks so much for taking the time to be involved.

[fallenwitch3r 41]

Okay, now I know it. There was some misunderstanding.

[fallenwitch3r 41]

2 minutes ago, chef said:

Im interested if you got your results with fast detect on or off? 

Thanks so much for taking the time to be involved.

Fast detection was on

[Painkiller8818 203]

Any idea when it will get released for all?

[chef 3745]

4 hours ago, Painkiller8818 said:

Any idea when it will get released for all?

 The plugin or the actual skip button?

[fbrassin 35]

I need a help to understand how to test.

I tried to make scan to all my library with fast detect.

For some shows intro has been fuond, but for others no intro has been found even if there is.

Then i tried to reset one of these season (with no intro detected) detection and run again "audio fingerprint" and "title sequence detection" but with fast detect disabled. The result didn't change.

For that season i have start and end intro set to zero, and so also "end credit start"

I have many series in this situation.

Shall i try anything else, or how to give info about my situation to help testing?

Tnx

[rbjtech 4257]

3 hours ago, fbrassin said:

I need a help to understand how to test.

I tried to make scan to all my library with fast detect.

For some shows intro has been fuond, but for others no intro has been found even if there is.

Then i tried to reset one of these season (with no intro detected) detection and run again "audio fingerprint" and "title sequence detection" but with fast detect disabled. The result didn't change.

For that season i have start and end intro set to zero, and so also "end credit start"

I have many series in this situation.

Shall i try anything else, or how to give info about my situation to help testing?

Tnx

Thanks for the feedback.

Remember that the Intro's are found using Audio matching - so is there a possibility that while the Intro Video looks the same, the Audio/Music behind it is different per episode ?

If you're happy to post the name an example show, then we may be able to test ourselves.  If not, PM myself, @chef or @Cheesegeezer and we can privately test.

Under the hood, there are various 'tuning' parameters to make the detection more or less 'sensitive' so it's possible we may need to tweak those, but given our extensive testing thus far, we believe we have it set to be the most accurate.

Edited November 17, 2021 by rbjtech

[fbrassin 35]

2 examples are the series Dark (all 3 seasons) and Foundation.

Both of them have got an intro that to me seems the same for each episode, but none of them has been detected.

Is this enough for you or do you need other info?

 

Edited:

if this can help, i tried to play dark with plex and intro was found and skip.

Edited November 17, 2021 by fbrassin

[Cheesegeezer 3086]

14 minutes ago, fbrassin said:

2 examples are the series Dark (all 3 seasons) and Foundation.

Both of them have got an intro that to me seems the same for each episode, but none of them has been detected.

Is this enough for you or do you need other info?

 

How long is the intro for both series.

could you enable debug logging and then restart the server and then run FP and detection again and post the log. If you feel sensitive please PM, myself or @chef the log and we will destroy after reviewing.

 

thanks again for your help

[chef 3745]

20 minutes ago, fbrassin said:

2 examples are the series Dark (all 3 seasons) and Foundation.

Both of them have got an intro that to me seems the same for each episode, but none of them has been detected.

Is this enough for you or do you need other info?

 

Edited:

if this can help, i tried to play dark with plex and intro was found and skip.

I can test Foundation. I have been following that series. 

I haven't actually checked my results on it yet.

I will test it this morning.

 

[rbjtech 4257]

15 minutes ago, fbrassin said:

2 examples are the series Dark (all 3 seasons) and Foundation.

Both of them have got an intro that to me seems the same for each episode, but none of them has been detected.

Is this enough for you or do you need other info?

 

That's all we need - thanks.

So I have Dark myself - and all 3 seasons appear to be picked up perfectly for Intros - S1E10 credit is missing for some reason.

@chef / @Cheesegeezer - I'm wondering if the OP's shows are simply not being scanned - ie same issue we may have identified a few days ago ?

 

 

 

[chef 3745]

6 minutes ago, rbjtech said:

That's all we need - thanks.

So I have Dark myself - and all 3 seasons appear to be picked up perfectly for Intros - S1E10 credit is missing for some reason.

@chef / @Cheesegeezer - I'm wondering if the OP's shows are simply not being scanned - ie same issue we may have identified a few days ago ?

 

 

 

We had best push that release that fixed the issue from yesterday. I'll send cheese the build, with an updated rev.

Edited November 17, 2021 by chef

[chef 3745]

@fbrassin while the other thread is being updated. Here is a prelease v_0.0.1.22

This is the same as the one that will be updated in the other thread. I'll post it here for you to try in the meantime.

IntroSkip_v0.0.1.22.zip

[rbjtech 4257]

9 minutes ago, chef said:

@fbrassin while the other thread is being updated. Here is a prelease v_0.0.1.22

This is the same as the one that will be updated in the other thread. I'll post it here for you to try in the meantime.

IntroSkip_v0.0.1.22.zip 240.23 kB · 0 downloads

Thanks @chef 

@fbrassin - Just copy the DLL over the existing one, and restart emby.  Do NOT delete the titlesequence.db file !

Re-run the FP and Detect tasks - if it runs through very quickly, then this is not the issue, but if it stops part of the way through, then it is picking up items it missed during the FP and/or Detect.  Once complete, hopefully your results match mine. 

Edited November 17, 2021 by rbjtech

[fbrassin 35]

OK, thx.

I'll try in a few minutes and let you know.

[neik 837]

On 11/14/2021 at 1:30 PM, Cheesegeezer said:

it shouldn't do. as we are using the chroma print function and also the black detects, however, i will read up on the changelog for ffmpeg.

Thanks for letting us know.

 

I suppose there wasn't anything alarming in the changelog, right?

[chef 3745]

I was able to get data on foundation:

 

However, I did find an error happening, when calculating the end credits. It happens when the emby base Item doesn't have runtime tick values in the metedata.

I'm going to see if we can calculate the runtime ourselves if it doesn't exist yet, using ffmpeg.

 

this is why the data posted in the picture doesn't have a completed set of credit data.

Edited November 17, 2021 by chef

[chef 3745]

Upon further examination of that code, specifically dealing with nullable runtime tick values. 

I believe it is really important to ask users  to make sure they have their metadata refreshed, and totally scanned before attempting sequence detection.

We can handle null exceptions from missing metadata, however there won't be results for end credit data without a runtime value for the episode.

Unless we ignore credit sequences end times, all together. Which could also be a possibility. 

 

 

Edited November 17, 2021 by chef

[chef 3745]

After refreshing the metadata on foundation, and turning fast detect off. My results were absolutely perfect. 

 

You can see the importance of utilizing black frame detection here in the log. So fast detect being turned off.

The audio fingerprint was unable to detect where the credit sequence start time was for this episode. It was in fact the black frame detection encoding which was able to pinpoint the position of the end credits stating

 

2021-11-17 10:36:24.806 Debug Intro Skip: DETECTION Adjusting black frame detection for Foundation Season 1 Episode 3
2021-11-17 10:36:38.700 Debug Intro Skip: Foundation Season 1 Episode 3:
Credit sequence audio detection start: 00:00:00.
Black frame detected within contiguous regions: 00:48:35.1210000.
Moving sequence start to: 00:48:35.1210000.
2021-11-17 10:36:38.700 Debug Intro Skip: 
DETECTION processed 8 compared items for Foundation Season 1 Episode 3.
Best result: Foundation - Season 1 Episode 3 
TITLE SEQUENCE START: 00:00:00 
TITLE SEQUENCE END: 00:01:23 
CREDIT START: 00:48:35.1210000
CREDIT END: 00:49:39.5200000
CREDIT CONFIDENCE: 0.46
TITLE SEQUENCE CONFIDENCE: 1

Audio found the credits at 00:00:00. Black detection sees this as the  "impossible situation", and takes over the detection to calculate the proper start time. And succeeds .

Edited November 17, 2021 by chef

[Cheesegeezer 3086]

wow, i was busy at work and you guys have solved world peace and the climate issues.

Awesome

Edited November 17, 2021 by Cheesegeezer

[rbjtech 4257]

1 hour ago, chef said:

Upon further examination of that code, specifically dealing with nullable runtime tick values. 

I believe it is really important to ask users  to make sure they have their metadata refreshed, and totally scanned before attempting sequence detection.

We can handle null exceptions from missing metadata, however there won't be results for end credit data without a runtime value for the episode.

Unless we ignore credit sequences end times, all together. Which could also be a possibility. 

 

 

Any reason why this value would be missing in the first place ?

Forcing a metadata refresh just to get this value seems a tad extreme and may overwrite peoples changes if they have not locked the metadata - if it's just the show runtime, then could we not ask emby to get it via an API call ?

[chef 3745]

16 minutes ago, rbjtech said:

Any reason why this value would be missing in the first place ?

Forcing a metadata refresh just to get this value seems a tad extreme and may overwrite peoples changes if they have not locked the metadata - if it's just the show runtime, then could we not ask emby to get it via an API call ?

The value in the API is a nullable int. Without going to heavy into the differences between reference types, and value types, it's really that the item has been found by emby, but it hasn't had ffprobe run on it yet (or something similar). So the runtime hasn't been calculated

Integers can not be null because of how integers are pointed to in memory.  They are a value type. An integer must always have a value: 0, a negative number, or a max value (whatever c# max integer value is)

But, c# allows us to create a nullable types, (int?). This makes the integer allowed to be null.

When it is a nullable type, we have to check if it "HasValue" before using it. Otherwise there is a null reference exception.

 

Making sure the scans are complete will make sure that the runtime values have a value saved, and we can use it. 

Now that I think about it, I'm not sure either if we should force a metadata scan. 

But, we should definitely log the items without runtime values, and maybe even put them in our stats report, so the user knows to refresh the data, and try again.

 

 

Edited November 17, 2021 by chef