Autoskip doesn't seem to respect comskip padding

This originally started as a feature request for an option to control the behavior of automatic commercial skipping in relation to the section markers.

In that thread I learned of the configuration options available in comskip, of which the padding option seemed to fit the bill perfectly. Unfortunately though, after quite a bit of experimentation (detailed in the thread), it doesn't appear that the automatic commercial skip actually respects the padded markers. Comskip appears to do it's job as expected, but you can see via Edit commercials, playback, etc. that it just doesn't seem to respect the adjusted markers.

Am I just doing something stupid?

Any info would be appreciated-

Comskip will always be behind the 8-ball because the networks actively try to foil it in unending ways. My solution? No comskip, set my remote for 30sec >> and 7sec<<, and avoid all the frustration.

Generally 4-5 forwards work great with an occasional 1 rewind if I overshoot the commercials.

This is one reason I like to use Kodi with either the Channels DVR add-on or the PlexKodiConnect add-on to play back my recordings. It has advanced skip forward and back options that are user defined. You can press skip multiple times and tell it how far to skip depending on how many presses. Works very well.

Appears that Channels is using info from the comskip log and relying on its scoring system to mark commercials.
When you edit commercials, it does a GET dvr/files/file#/comskip/info?source=local&apply_edits=false
and appears to pull info from this block list and use the score (sc column) to determine commercial or program.
The block numbers correspond to the blocks you see in edit commercials.
The -- or ++ after the block number appears to mean --(commercial) and ++(program).

Score threshold:            1.0500
~
----------------------------------------------------
Block list after weighing
----------------------------------------------------
  #     sbf  bs  be     fs     fe        ts        te       len     sc   scr cmb   ar                   cut    bri logo   vol sil   corr stdev   cc
  0:--    4   0   4      1   1016     0.00s    16.95s    16.95s   1.50  0.06   0 1.78                  v   b   539- 0.00  409- 35-  0.000    53 NONE      
  1:++    4   4   0   1017   8273    16.97s   138.02s   121.05s   0.01  0.30   0 1.78             L    v      1973+ 0.49  683- 61+  0.000   197 POPON     
~
 77:--    1   1   0 196200 198844  3273.32s  3317.43s    44.11s   1.50  0.11   0 1.78                  v   b   837+ 0.00  815+ 84+  0.000    83 NONE      
 78:++    1   1   1 198845 216884  3317.45s  3618.40s   300.95s   0.00  0.07   0 1.78            EL    v       615- 0.66  615- 49-  0.000    61 POPON     
 79:++    1   1   0 216885 217515  3618.41s  3628.91s    10.49s   1.00  0.02   0 1.78  7      {        v a br  442- 0.00  661- 56-  0.000    44 POPON     
1 Like

That is very interesting. If I'm understanding you then, it doesn't sound like there's much I can do as it truly isn't respecting the adjusted comskip markers but rather focuses on the overall blocks?

I appreciate you looking into it. What are the chances of one of the devs seeing this and weighing in?

Thank you!

Chances are high they've seen it. Whether they adjust their internal algorithms to focus upon a different point than they already do is up in the air. However, since the discrepancy has been shown, and its effect, chances are higher for change than you might expect.

1 Like

Thanks everyone for the feedback! Maybe the devs will take an interest in it! :slight_smile:

Looks like for recordings using the new Smart Comerical detect it does a GET /dvr/files/file#/comskip/info to get the commercial/program blocks and relies on the boolean value true or false in Blocks.IsCommercial