Audio fingerprinting for old files?

[zodac01 3]

Hi, 

Is it normal for Emby fmto generate fresh audio fingerprints for TV shows that have been scanned/imported before? For example, I have old seasons of multiple that were imported long ago (some I have watched since importing, some not), and have had their 'skip intro' detection completed. 

However, almost every day when I check my dashboard I see my library scan hanging at ~94% as it generates fresh audio fingerprints. 

Is this normal, or is there some way to disable fingerprinting if one exists? 

Thanks! 

[Luke 42079]

Hi there, why do you think it's generating fresh fingerprints?

[rbjtech 5284]

see - https://emby.media/community/index.php?/topic/120197-core-skip-intro-new-episodes-do-re-scan-of-all-episodes/#comment-1268355

@zodac01- A debug log of the events will show what's going on. 

[zodac01 3]

On 24/07/2023 at 18:29, Luke said:

Hi there, why do you think it's generating fresh fingerprints?

 

I was seeing the following logs  each day when I did a library scan, for the same show/season:
 

2023-07-26 15:38:31.913 Info App: Best title sequence for B:\TV\Show\Season 1\S01E01.mkv is 00:00:29.4776119 - 00:01:58.5323382. Confidence: 0.93
2023-07-26 15:38:34.510 Info App: Best title sequence for B:\TV\Show\Season 1\S01E02.mkv is 00:00:30.2238805 - 00:01:57.1641790. Confidence: 1
2023-07-26 15:38:37.114 Info App: Best title sequence for B:\TV\Show\Season 1\S01E03.mkv is 00:08:43.1343283 - 00:09:01.0447760. Confidence: 0.9

My understanding was if a time window + confidence level was being logged, it was generating a fresh fingerprint?

On 24/07/2023 at 21:42, rbjtech said:

see - https://emby.media/community/index.php?/topic/120197-core-skip-intro-new-episodes-do-re-scan-of-all-episodes/#comment-1268355

@zodac01- A debug log of the events will show what's going on. 

 

This actually helped me figure out what was going on - I guess the fingerprinting occurs for a full season if/when new episodes are added? I had some corrupted files from an old restore, and they were spread across a few different shows/seasons. I suppose when the ffprobe failed, it would still attempt to extract info from these files on subsequent scans, and therefore keep generating fingerprints for the season? The error I was getting was:
 

ffprobe version 5.0.0-emby_2022_05_27-u1 Copyright (c) 2007-2022 the FFmpeg developers and softworkz for Emby LLC
	  built with gcc 10.3.0 (Rev5, Built by MSYS2 project)
	[matroska,webm @ 0000022328760840] Format matroska,webm detected only with low score of 1, misdetection possible!
	[matroska,webm @ 0000022328760840] EBML header parsing failed
	file:A:\TV\Show\Season 1\S01E04.mkv: Invalid data found when processing input
	
	
	   at Emby.Server.MediaEncoding.Probing.MediaProbeManager.GetMediaInfoInternal(String inputPath, String primaryPath, MediaProtocol protocol, String userAgent, Boolean extractChapters, String probeSizeArgument, Boolean isAudio, Boolean forceEnableLogging, CancellationToken cancellationToken)
	   at Emby.Providers.MediaInfo.FFProbeVideoInfo.ProbeVideo[T](MetadataResult`1 metadataResult, MetadataRefreshOptions options, LibraryOptions libraryOptions, CancellationToken cancellationToken)
	   at Emby.Providers.Manager.MetadataService`2.RunCustomProvider(ICustomMetadataProvider`1 provider, MetadataResult`1 itemResult, String logName, MetadataRefreshOptions options, LibraryOptions libraryOptions, RefreshResult refreshResult, CancellationToken cancellationToken)
	Source: Emby.Server.MediaEncoding
	TargetSite: Void MoveNext()

I've since replaced/removed the corrupted files, and after a fresh run of generating fingerprints for the impacted seasons, things look to be running smoothly again.

[rbjtech 5284]

9 hours ago, zodac01 said:

 

I was seeing the following logs  each day when I did a library scan, for the same show/season:
 

2023-07-26 15:38:31.913 Info App: Best title sequence for B:\TV\Show\Season 1\S01E01.mkv is 00:00:29.4776119 - 00:01:58.5323382. Confidence: 0.93
2023-07-26 15:38:34.510 Info App: Best title sequence for B:\TV\Show\Season 1\S01E02.mkv is 00:00:30.2238805 - 00:01:57.1641790. Confidence: 1
2023-07-26 15:38:37.114 Info App: Best title sequence for B:\TV\Show\Season 1\S01E03.mkv is 00:08:43.1343283 - 00:09:01.0447760. Confidence: 0.9

My understanding was if a time window + confidence level was being logged, it was generating a fresh fingerprint?

 

This actually helped me figure out what was going on - I guess the fingerprinting occurs for a full season if/when new episodes are added? I had some corrupted files from an old restore, and they were spread across a few different shows/seasons. I suppose when the ffprobe failed, it would still attempt to extract info from these files on subsequent scans, and therefore keep generating fingerprints for the season? The error I was getting was:
 

ffprobe version 5.0.0-emby_2022_05_27-u1 Copyright (c) 2007-2022 the FFmpeg developers and softworkz for Emby LLC
	  built with gcc 10.3.0 (Rev5, Built by MSYS2 project)
	[matroska,webm @ 0000022328760840] Format matroska,webm detected only with low score of 1, misdetection possible!
	[matroska,webm @ 0000022328760840] EBML header parsing failed
	file:A:\TV\Show\Season 1\S01E04.mkv: Invalid data found when processing input
	
	
	   at Emby.Server.MediaEncoding.Probing.MediaProbeManager.GetMediaInfoInternal(String inputPath, String primaryPath, MediaProtocol protocol, String userAgent, Boolean extractChapters, String probeSizeArgument, Boolean isAudio, Boolean forceEnableLogging, CancellationToken cancellationToken)
	   at Emby.Providers.MediaInfo.FFProbeVideoInfo.ProbeVideo[T](MetadataResult`1 metadataResult, MetadataRefreshOptions options, LibraryOptions libraryOptions, CancellationToken cancellationToken)
	   at Emby.Providers.Manager.MetadataService`2.RunCustomProvider(ICustomMetadataProvider`1 provider, MetadataResult`1 itemResult, String logName, MetadataRefreshOptions options, LibraryOptions libraryOptions, RefreshResult refreshResult, CancellationToken cancellationToken)
	Source: Emby.Server.MediaEncoding
	TargetSite: Void MoveNext()

I've since replaced/removed the corrupted files, and after a fresh run of generating fingerprints for the impacted seasons, things look to be running smoothly again.

Good analysis - and I suspect that is exactly what it is doing. 

Remember, Chromaprint requires all the existing fingerprints from that Season to compare against (that's what the 'confidence' score is - it's a calculation.)  If a FP doesn't exist, then it will try and generate it each time, thus keep looping.