Plugin: EmbyCredits, detect end credits and add auto skip.

[Ferwin29 13]

yes

I restarted twice to be sure.

Edited Wednesday at 04:10 PM by Ferwin29

[yocker 1974]

1 hour ago, Ferwin29 said:

Sorry, I've got the same error in 2.9.0.3

Try version 2.9.0.5 beta in the catalog.

[Ferwin29 13]

It's ok now, thank you  

[yocker 1974]

Just now, Ferwin29 said:

It's ok now, thank you  

No problem.  

[fbourg 8]

Hello @yocker,

I think the problem was actually on my end. I had an Unraid plugin that stops all containers overnight for backups, sorry about that! That said, I had no issues at all last night.

[yocker 1974]

3 hours ago, fbourg said:

Hello @yocker,

I think the problem was actually on my end. I had an Unraid plugin that stops all containers overnight for backups, sorry about that! That said, I had no issues at all last night.

That would explain it! Been scratching my head hairless figuring out what it could be. 

Happy to hear it's working now!  

[fbourg 8]

3 minutes ago, yocker said:

That would explain it! Been scratching my head hairless figuring out what it could be. 

Happy to hear it's working now!  

Really sorry for the false alarm and for making you scratch your head over this!

Thanks a lot for your quick replies and for this great plugin. Everything is running smoothly now!

[yocker 1974]

1 minute ago, fbourg said:

Really sorry for the false alarm and for making you scratch your head over this!

Thanks a lot for your quick replies and for this great plugin. Everything is running smoothly now!

No need to apologize.  
No problem and thank you too!  

[sydlexius 327]

9 hours ago, fbourg said:

Hello @yocker,

I think the problem was actually on my end. I had an Unraid plugin that stops all containers overnight for backups, sorry about that! That said, I had no issues at all last night.

@fbourgSomething you can consider (even if temporary) is to set the "backup/restore appdata" plugin to skip Emby; you can also potentially replace it with Emby's own "Backup" plugin and avoid downtime altogether. With that said, I do the same thing that you do, as I have found restoring from the appdata backup to be quite a bit more seamless.

Edited Thursday at 03:49 PM by sydlexius

[fbourg 8]

@yockerDo you know why I have nearly 17,000 skipped episodes? I don't have the log file yet, should I enable OCR?

@sydlexiusthank you

[yocker 1974]

1 hour ago, fbourg said:

@yockerDo you know why I have nearly 17,000 skipped episodes? I don't have the log file yet, should I enable OCR?

@sydlexiusthank you

Chroma/Hash detection needs either the same music between episodes or a black scene transition.
If none of those are in the episodes then the detection will fail.

A propper set up OCR can help with that, yes.

[fbourg 8]

7 minutes ago, yocker said:

Chroma/Hash detection needs either the same music between episodes or a black scene transition.
If none of those are in the episodes then the detection will fail.

A propper set up OCR can help with that, yes.

Thanks for clarifying! Just to be sure I understand correctly: are the ~17,000 'skipped' episodes actually considered successful because the timings were resolved through batch audio comparison (since it says 'Only needed for audio comparison'), meaning only the ~4,600 'failed' episodes need OCR fallback?

[yocker 1974]

26 minutes ago, fbourg said:

Thanks for clarifying! Just to be sure I understand correctly: are the ~17,000 'skipped' episodes actually considered successful because the timings were resolved through batch audio comparison (since it says 'Only needed for audio comparison'), meaning only the ~4,600 'failed' episodes need OCR fallback?

I should remove that from being shown as they are not truly "skipped".

Chroma needs other episodes to compare the sound to before it knows what is the credit transition and there for what to look for. The skipped you see are just episodes it used for the comparison and not episodes it actually ran the detection on hence the "Only needed for audio comparison".

There are also other reasons episodes are skipped, like for example they already have been run though detection, they are configured to be skipped in the rules sections and/or they failed previously (if that setting has been enabled).
They are simply skipped to save time and not having to re-run detection every time on them when not needed.

Quick disclaimer.
The plugin's detection is not perfect, it needs episodes to either have the same music or a black scene transition. If none of those are there it will fail.
Using OCR can help with that as almost all TV shows use credit scenes with text. As long as the OCR is set to look for the correct keywords and the text is not some strange fuzzy or stylized text, like anime tend to have, it will find the credits.
This requires a bigger computer though as it's basically a small AI that is used to read the text.

While OCR works and is in general more precise than the other detections i don't use it my self, the small Dell machine i have for my Emby is too small for it.
Only my test and main machine have the horse power to run it.

TL:DR.
If you feel like there are too many failed episodes you can try and run manual detection on them again one TV show at a time.
If you think they shouldn't have failed then send me a log from the "dry run with debug" and i will look at it. (One TV show at a time please so i don't get a 50MB log  ). 

Edited 12 hours ago by yocker

[fbourg 8]

 

1 hour ago, yocker said:

I should remove that from being shown as they are not truly "skipped".

Chroma needs other episodes to compare the sound to before it knows what is the credit transition and there for what to look for. The skipped you see are just episodes it used for the comparison and not episodes it actually ran the detection on hence the "Only needed for audio comparison".

There are also other reasons episodes are skipped, like for example they already have been run though detection, they are configured to be skipped in the rules sections and/or they failed previously (if that setting has been enabled).
They are simply skipped to save time and not having to re-run detection every time on them when not needed.

Quick disclaimer.
The plugin's detection is not perfect, it needs episodes to either have the same music or a black scene transition. If none of those are there it will fail.
Using OCR can help with that as almost all TV shows use credit scenes with text. As long as the OCR is set to look for the correct keywords and the text is not some strange fuzzy or stylized text, like anime tend to have, it will find the credits.
This requires a bigger computer though as it's basically a small AI that is used to read the text.

While OCR works and is in general more precise than the other detections i don't use it my self, the small Dell machine i have for my Emby is too small for it.
Only my test and main machine have the horse power to run it.

TL:DR.
If you feel like there are too many failed episodes you can try and run manual detection on them again one TV show at a time.
If you think they shouldn't have failed then send me a log from the "dry run with debug" and i will look at it. (One TV show at a time please so i don't get a 50MB log  ). 

Thanks for the explanation, that makes total sense! Yes, removing or clarifying that 'skipped' display would definitely prevent confusion. I'll let the scan finish and then test manual OCR detection on the failed shows.

[yocker 1974]

32 minutes ago, fbourg said:

 

Thanks for the explanation, that makes total sense! Yes, removing or clarifying that 'skipped' display would definitely prevent confusion. I'll let the scan finish and then test manual OCR detection on the failed shows.

No problem. 

Tbh. the default settings might be a little too tight as well. Made them like that to make them as precise as possible and limit the number of false positives. 

It's an on going "battle" to find settings that are juuuuust right.  

[Baenwort 125]

I view this as doing it it layers. I want to start off with tight settings and then it would be nice to have a button for 'looser' settings so you could easily swap between pre-sets depending on how much it bothers you to have the marker wrong.

I am seeing an issue with 2.9.0.5 in that it seems to stop with a 'complete' mark after only doing a couple hundred episodes even though it says there are over 10K left to do.

I'll reboot the logs and try to minimize activity to get a clean capture tonight after everyone goes to bed.

[yocker 1974]

1 hour ago, Baenwort said:

I view this as doing it it layers. I want to start off with tight settings and then it would be nice to have a button for 'looser' settings so you could easily swap between pre-sets depending on how much it bothers you to have the marker wrong.

I am seeing an issue with 2.9.0.5 in that it seems to stop with a 'complete' mark after only doing a couple hundred episodes even though it says there are over 10K left to do.

I'll reboot the logs and try to minimize activity to get a clean capture tonight after everyone goes to bed.

Right now it counts all episodes in the queue, even those that doesn't actually need any detection.
So as the queue goes on it might find episodes that it didn't need to run detection on and there for not counted.

I'm not sure precisely to do about it but i am working on it.

[Baenwort 125]

 

When I press the queue button it says there are no pending but as soon as I set to process (and I've set it to not retry failed detections) it always finds more to do and processes a couple hundred episodes before showing 'complete' but numerically having more to go.

[yocker 1974]

@Baenwort
I made a small change to how skipped files are counted towards progress and how some failed files were not marked right.

The change are in the 2.9.1.0 beta in the catalog.

[yocker 1974]

1 hour ago, Baenwort said:

 

When I press the queue button it says there are no pending but as soon as I set to process (and I've set it to not retry failed detections) it always finds more to do and processes a couple hundred episodes before showing 'complete' but numerically having more to go.

This might be fixed in the version i just wrote about.