Inaccurate commercial skip

[jtanderson]

I've been seeing inaccurate commercial skips with both the Windows client and the NextPVR Kodi add-on.  It appears this problem may be occurring only with 480i content and it is really a problem with timeline calculation.  In general, NextPVR is both starting and ending the skip prematurely.

I have verified that the skip points generated by comskip are accurate using both the review function of comskip and by verifying the edl file using VideoReDo.

In one example, the edl file shows the beginning of a commercial break at 560.73 seconds (~9:20).  The commercial marker indeed starts at 9:20 according to the NextPVR client timeline.  But the 9:20 point on the timeline isn't really the 9:20 point in the program.  The point in the stream that the NextPVR client shows as 9:20 is really the 8:22 point in the stream according to VideoReDo.  The actual commercial break starts at the 10:17 mark according to the NextPVR client timeline.

The program in question is 480i, 29.97 fps.  It was recorded from an OTA digital sub-channel from an HDHomeRun.  The end of the program is at 29:50 according to the NextPVR timeline and 30:02 according to VideoReDo.

I checked out a 1080i program and a 720p program and in both cases the timeline was acceptably accurate.  I've seen this problem with every 480i program I have looked at, but that is admittedly a small sample.

[mvallevand]

Comskip and VideoRedo will be able to calculate timelines much more accurately because they aren't doing it on the fly and they read the actual mpeg stream and not the ts container.

Post your Kodi debug log playing the file and I can check out the Kodi side.

If you post a link to that file somewhere with the EDL file I can have a look with Kodi from my side if you want.

Martin

[jtanderson]

It may take me a while to get around to this. I would actually prefer to focus on using the NextPVR client as that reduces the number of variables.

[mvallevand]

Ok, I expect the issue is the file itself and the Kodi debug logs are more detailed so if it experiences the same problem it might be easier to find.

In general this is not the type of problem I'd expect to solve without the sample file.

Martin