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

[yocker 1947]

4 minutes ago, fbourg said:

High failure rate despite OCR being enabled

That points to the problem being the files if all detections fail.
Can i get you to run a dry run with debug on one of those TV shows that there is a problem with and upload the log file that it gives?
Like for example the "Jane the Virgin" 

[yocker 1947]

5 hours ago, fbourg said:

I haven't tested this version yet. I started it last night on 8.1, but when I woke up, the scan had stopped again. Even after adjusting the OCR, I still have quite a few episodes failing, even with 'OCR with hash fallback' enabled.

embyserver.txt 4.75 MB · 0 downloads

I'll look into it.
My theory as it is right now is that you might have some other plugin or outside program that might interfere with it as i have recently added automatic restarts if the plugin ever stops before the detections are done. 

Edit: Scratch that, i might know what's wrong. Will start on a fix now.

Edited 11 hours ago by yocker

[fbourg 4]

log file for jane

credits-detection-debug-2026-09-21T12-53-19.log

[yocker 1947]

18 minutes ago, fbourg said:

log file for jane

credits-detection-debug-2026-09-21T12-53-19.log 27.77 kB · 0 downloads

Please run it for a whole season and not just one file.

Edit: Mistyped, that was not for you, sorry.. Doing multiple things at once right now, confusing as hell. 

Edited 9 hours ago by yocker

[yocker 1947]

30 minutes ago, fbourg said:

log file for jane

credits-detection-debug-2026-09-21T12-53-19.log 27.77 kB · 0 downloads

Think i see the problem.
Will have a fix in next version.

[fbourg 4]

new file with season

credits-detection-debug-2026-09-21T13-27-02.log

[fbourg 4]

2.8.9.3 is new version ? 

[yocker 1947]

20 minutes ago, fbourg said:

2.8.9.3 is new version ? 

Will be in the next one after that.
This was just a small quick fix to the black frame detection fallback.

[yocker 1947]

3 hours ago, fbourg said:

2.8.9.3 is new version ? 

I just uploaded 2.8.9.4. See if that one helps.

[fbourg 4]

1 hour ago, yocker said:

I just uploaded 2.8.9.4. See if that one helps.

same problem with severance (i tested only this show)

credits-detection-debug-2026-09-21T21-00-21.log

[fbourg 4]

1 hour ago, yocker said:

I just uploaded 2.8.9.4. See if that one helps.

Time too longer (3000 hours)

[yocker 1947]

35 minutes ago, fbourg said:

same problem with severance (i tested only this show)

credits-detection-debug-2026-09-21T21-00-21.log 246.43 kB · 1 download

That should have fixed the:
[swscaler @ 0x1feb500] [swscaler @ 0x22632c0] deprecated pixel format used, make sure you did set range correctly

Will keep trying.

Can you try and do a short detection with CPU and then Hardware acceleration only if possible?

Edited 1 hour ago by yocker

[yocker 1947]

15 hours ago, fbourg said:

I went back to the default settings (I had added +fra to the OCR as well as quite a few keywords; now I've removed both fra and jpn, along with the Japanese keywords, and it's running much better. I'll add them back once I've finished the essentials). That said, the scan could still stop again at any moment.

I just re-read this.
The keywords does not support jpn and fra as in the language codes. The OCR engine only looks for letters so adding a language code will do nothing.
Just add keywords that you know are in the credits of the episodes like for example "written", "story" "actor" and so on.

[fbourg 4]

Thanks for the clarification on the keywords and language codes. Just to confirm, the latest tests were already done with "eng" only—I had already removed both "fra" and "jpn" as well as the non-English keywords.

  

Regarding the range warning (make sure you did set range correctly), I wasn't sure if that was a setting I had to configure on my end or just an internal FFmpeg warning.

  

Those tests were run with QSV. I will run a short detection using CPU only as requested, test it again with hardware acceleration, and share the new logs.

[yocker 1947]

14 minutes ago, fbourg said:

Thanks for the clarification on the keywords and language codes. Just to confirm, the latest tests were already done with "eng" only—I had already removed both "fra" and "jpn" as well as the non-English keywords.

  

Regarding the range warning (make sure you did set range correctly), I wasn't sure if that was a setting I had to configure on my end or just an internal FFmpeg warning.

  

Those tests were run with QSV. I will run a short detection using CPU only as requested, test it again with hardware acceleration, and share the new logs.

The range setting is nothing you can do anything about. It's just noise indirectly caused by the yuv420p format of the video.
I should remove that from being logged.