1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

Sharing commercial skip files

Discussion in 'TiVo Home Media Features & TiVoToGo' started by moyekj, Mar 5, 2016.

  1. Mar 5, 2016 #1 of 60
    moyekj

    moyekj Well-Known Member

    11,298
    100
    Jan 23, 2006
    Mission...
    With the introduction of kmttg AutoSkip which allows you to automatically skip commercials when viewing shows on series 4 or later TiVos the first step towards being able to share commercial skip files is in place.

    Of course the problem is we have to currently manually generate our own .edl or .Vprj skip files for this to work. It would be nice to have a database of skip files available for download so that the community could share skip files with each other. That's what this thread is about - trying to come up with a viable scheme to allow sharing of commercial skip files.

    TiVo with introduction of SkipMode is effectively doing that. They have a database of skip files. They generate a skip file on a specific TiVo and then came up with a way to share that data with other TiVos. Of course the problem is that the precise cut points can vary from TiVo to TiVo, so there needs to be a method to adjust cut points to a particular TiVo, otherwise it's worthless.

    TiVo came up with a method to make those adjustments by using closed caption data which is assumed to be universally available among the TiVos sharing skip data. If you know the precise location in time of captions located close to skip points, you can adjust the skip points accordingly to account for timing differences between TiVos and thus effectively share skip data.

    In light of the above, I was thinking we could mimic the TiVo mechanism to some degree. If we collect 2 sets of files for a particular TiVo recording of a show:
    1. A .edl or .Vprj file with skip points
    2. A .srt file containing caption data for that same recording

    Then we can have some code that looks at the .srt file and the cut point file to determine time offsets of caption lines surrounding cut points. i.e. Attempting to depict this:

    TIVO 1
    captions_1-----cut point-----captions_2

    TIVO 2
    ---------captions_1-----cut point------captions_2

    If we can assume the time offset between captions_1 or captions_2 and cut point is the same for both TiVos then we can precisely adjust cut point for TIVO 2 using data from TIVO 1 and .srt data for TIVO 2. Multiple caption lines would have to be used to try and figure things out obviously since a single caption line by itself will not be unique for many cases.

    Note that the above is just brainstorming. I have no idea how practical it is to implement. But I'd like to explore it further or perhaps get ideas from others on more effective ways of sharing skip data.

    Perhaps as a 1st step assuming we pursue the above is to pick a certain show and generate skip files and .srt files for it and share the files with each other. Then just manually examine the files and see if we can manually generate an adjusted skip file that would work on our TiVo. Anybody up for doing such an experiment?
     
    Last edited: Mar 5, 2016
  2. Mar 5, 2016 #2 of 60
    lew

    lew Active Member

    4,009
    9
    Mar 12, 2002
    I'm not sure how often I'll be using kmttg autoskip however I'd be willing to participate in a test. Some people will be very precise with their cut points, others will be a second or two off.
     
  3. Mar 5, 2016 #3 of 60
    moyekj

    moyekj Well-Known Member

    11,298
    100
    Jan 23, 2006
    Mission...
    We need to find a show we both have recordings for. Here's some recent ones I have:
    Grimm [Ep 511] - Key Move
    Elementary [Ep 415] - Up to Heaven and Down to Hell
    Colony [Ep 108] - In From the Cold
    The 100 [Ep 307] - Thirteen
    The Family [Ep 101] - Pilot
    Marvel's Agent Carter [Ep 210] - Hollywood Ending
    Blindspot [Ep 111] - Cease Forcing Enemy
    Hawaii Five-0 [Ep 616] - Ka Pohaku Kihi Pa'a
    Marvel's Agent Carter [Ep 208] - The Edge of Mystery; A Little Song and Dance
    Hawaii Five-0 [Ep 615] - Ke Koa Lokomaika'i
    Grimm [Ep 510] - Map of the Seven Knights
    Marvel's Agent Carter [Ep 206] - Life of the Party; Monsters
    Grimm [Ep 509] - Star-Crossed
     
  4. Mar 5, 2016 #4 of 60
    reneg

    reneg Member

    720
    3
    Jun 19, 2002
    If needed, I can help out. I have recordings of the following from your list:

     
  5. Mar 5, 2016 #5 of 60
    moyekj

    moyekj Well-Known Member

    11,298
    100
    Jan 23, 2006
    Mission...
    OK, great. How about Blindspot Ep 111 then? I'll generate .Vprj file with my cut points and .srt file for it and put it in a zip file shortly.

    NOTE: Use TS container when downloading so as to have less chance of captions in .srt file being corrupted.
     
  6. Mar 5, 2016 #6 of 60
    moyekj

    moyekj Well-Known Member

    11,298
    100
    Jan 23, 2006
    Mission...
    OK here's my data for Blindspot. Note that I didn't mark the lead in commercial or the trailing commercial since I normally don't bother with those. From kmttg AutoSkip table these are the reported cut points:

    AutoSkip data for 'Blindspot - Cease Forcing Enemy'
    0: start=00:00 end=16:18
    1: start=18:50 end=26:18
    2: start=29:55 end=34:07
    3: start=38:25 end=43:30
    4: start=47:32 end=59:56

    And more precisely from the AutoSkip.ini file in msecs:
    0 978144
    1130562 1578977
    1795460 2047178
    2305670 2610541
    2852416 3596000

    For the above points here's closest meaningful CC data from my .srt file:
    Before "end=16:18"
    343
    00:16:12,072 --> 00:16:13,237
    The passengers.

    344
    00:16:13,239 --> 00:16:14,772
    They're still alive.

    After "start=18:50"
    347
    00:18:54,634 --> 00:18:55,967
    Do you speak English?

    348
    00:18:55,969 --> 00:18:58,302
    I do.
    My name is Kurt Weller.

    I think for now just focusing on those 2 points would be enough to get started with the analysis.

    EDIT: Looking at above data with all timing converted to msecs here's what I get:
    CC 343: 00:16:12,072 --> 00:16:13,237
    972072 - 973237 => -6072 to -4907 away from my end point

    CC 344: 00:16:13,239 --> 00:16:14,772
    973239 - 974772 => -4905 to -3372 away from my end point

    CC 347: 00:18:54,634 --> 00:18:55,967
    1134634 - 1135967 => 4072 to 5405 away from my start point

    CC 348: 00:18:55,969 --> 00:18:58,302
    1135969 - 1138302 => 5407 to 7740 away from my start point

    So if someone else posts their .srt file we can compute the location of end point and start point relative to their CC data and see if those points are close to good cut points for their recording.
     

    Attached Files:

    Last edited: Mar 5, 2016
  7. Mar 5, 2016 #7 of 60
    reneg

    reneg Member

    720
    3
    Jun 19, 2002
    My data for Blindspot. I ran combined decrypt & qsfix on the file before running ad detect. My vPrj cut points are on I-frames. I extracted captions without running ad-cut.

    Vprj Data:
    Code:
    <ProjectTime>25226226201</ProjectTime>
    <CutList><cut Sequence="1" CutStart="00:00:00;00" CutEnd="00:00:09;15" Elapsed="00:00:00;00"><CutTimeStart>0</CutTimeStart>
    <CutTimeEnd>95095112</CutTimeEnd>
    <CutByteStart>0</CutByteStart>
    <CutByteEnd>15373133</CutByteEnd>
    </cut><cut Sequence="2" CutStart="00:16:20;18" CutEnd="00:18:52;04" Elapsed="00:16:11;03"><CutTimeStart>9805796112</CutTimeStart>
    <CutTimeEnd>11321310112</CutTimeEnd>
    <CutByteStart>1613090757</CutByteStart>
    <CutByteEnd>1862538557</CutByteEnd>
    </cut><cut Sequence="3" CutStart="00:26:21;21" CutEnd="00:29:57;03" Elapsed="00:23:40;18"><CutTimeStart>15816801112</CutTimeStart>
    <CutTimeEnd>17970953112</CutTimeEnd>
    <CutByteStart>2602087881</CutByteStart>
    <CutByteEnd>2956524093</CutByteEnd>
    </cut><cut Sequence="4" CutStart="00:34:09;26" CutEnd="00:38:27;19" Elapsed="00:27:53;11"><CutTimeStart>20498478112</CutTimeStart>
    <CutTimeEnd>23076053112</CutTimeEnd>
    <CutByteStart>3372372949</CutByteStart>
    <CutByteEnd>3796463913</CutByteEnd>
    </cut><cut Sequence="5" CutStart="00:43:32;06" CutEnd="00:47:34;02" Elapsed="00:32:57;28"><CutTimeStart>26122096112</CutTimeStart>
    <CutTimeEnd>28540512112</CutTimeEnd>
    <CutByteStart>4297689961</CutByteStart>
    <CutByteEnd>4695676937</CutByteEnd>
    </cut><cut Sequence="6" CutStart="00:56:02;12" CutEnd="00:59:04;03" Elapsed="00:41:26;08"><CutTimeStart>33623590112</CutTimeStart>
    <CutTimeEnd>35440405112</CutTimeEnd>
    <CutByteStart>5531852997</CutByteStart>
    <CutByteEnd>5830770929</CutByteEnd>
    </cut><cut Sequence="7" CutStart="00:59:40;12" CutEnd="01:00:00;27" Elapsed="00:42:02;19"><CutTimeStart>35803768112</CutTimeStart>
    <CutTimeEnd>36008973112</CutTimeEnd>
    <CutByteStart>5890619225</CutByteStart>
    <CutByteEnd>5924652868</CutByteEnd>
    </cut></CutList>
    Autoskip.ini data:
    Code:
    <entry>
    contentId=tivo:ct.340036130
    offerId=tivo:of.ctd.3929992.612.cable.2016-03-01-03-00-00.3600
    offset=0
    tivoName=Living Room
    title=Blindspot - Cease Forcing Enemy
    0 0
    9509 980579
    1132131 1581680
    1797095 2049847
    2307605 2612209
    2854051 3362359
    3544040 3580376
    3600897 3594000
    
    SRT file near first commercial:
    Code:
    342
    00:16:15,040 --> 00:16:16,207
            The passengers.         
    
    343
    00:16:16,209 --> 00:16:17,742
     They're still alive.           
    
    344
    00:18:55,400 --> 00:18:57,602
                   ΒΆ                
    
    345
    00:18:57,604 --> 00:18:58,936
         Do you speak English?      
    
    
     

    Attached Files:

  8. Mar 5, 2016 #8 of 60
    moyekj

    moyekj Well-Known Member

    11,298
    100
    Jan 23, 2006
    Mission...
    OK, so based on your srt data and applying my offsets I get the following assuming I did the math correctly:
    343
    00:16:16,209 --> 00:16:17,742
    976209 - 977742 => end point = 977742 + 3372 = 981114 = 16:21 (980579) => 535 msec difference vs yours

    345
    00:18:57,604 --> 00:18:58,936
    1137604 - 1138936 => start point = 1137604 - 4072 = 1133532 18:54 (1132131) => 1401 msec difference vs yours

    So if you can check that the following would be viable for your Blindspot recording:
    981114 msecs or ~ 16:21 = start of 1st commercial
    1133532 msecs or ~ 18:54 = end of 1st commercial

    EDIT: The reverse analysis using your points and applying your CC offsets to my CC data I get very close to what I had marked:

    your end point: 980579
    your closest cc point: 977742 => -2837 away from end point

    your start point: 1132131
    your closest cc point: 1137604 => 5473 away from start point

    => end point: 974772 + 2837 = 977609 =~ 16:18 (I marked ~ 16:18)

    => start point: 1134634 - 5473 = 1129161 =~ 18:49 (I marked ~ 18:50)
     
    Last edited: Mar 5, 2016
  9. Mar 5, 2016 #9 of 60
    reneg

    reneg Member

    720
    3
    Jun 19, 2002
    Using VideoRedo to check:
    16:21:00 would be viable, It is a black screen.
    I-frame
    16:20.18 - Black video (My start of the 1st commercial cutpoint)
    16:22.06 - Start of 1st commercial


    At 18:54:00, it is into the show.
    I- Frame
    18:52.04 - black screen (my end of 1st commercial cutpoint)
    18:52.19 - Start of show after commercial
    18:53.19
    18:54.19 - Nearest I-frame to 18:54

    Could the commercial on your system have run long? I have an additional caption at
    Code:
    344
    00:18:55,400 --> 00:18:57,602            
    
    which was not in your SRT
     
  10. moyekj

    moyekj Well-Known Member

    11,298
    100
    Jan 23, 2006
    Mission...
    I think mine is there but I didn't use it for the analysis since it's not text:
    346
    00:18:51,163 --> 00:18:52,430
    &#9834;&#9834;

    Let me double-check that my end of 1st commercial point is indeed valid for my recording. EDIT: Yup, my point is perfect for my recording - a black frame right before the show starts again.

    Perhaps you can double-check my math as could be I screwed it up for that point. Note that it's 18:53.5 the actual point so probably about 1 second into your show.
     
    Last edited: Mar 5, 2016
  11. moyekj

    moyekj Well-Known Member

    11,298
    100
    Jan 23, 2006
    Mission...
    We should probably look at the other cut points as well to get a complete picture how accurate this can be expected to be... Right now I'm doing all this manually which is pretty painful, but I don't want to spend time writing a program just yet if it turns out it's not going to work out very well. I will note that using TiVo SKIP data sometimes it skips about 1 second or so into the show for me so their methodology is not perfect either, but I don't expect it to be perfect. As long as it's within 1 or 2 seconds at the most I'm OK with it.
     
  12. moyekj

    moyekj Well-Known Member

    11,298
    100
    Jan 23, 2006
    Mission...
    OK, did a full analysis with the help of some code using your data which I capture here:
    Code:
    9509 980579
    1132131 1581680
    1797095 2049847
    2307605 2612209
    2854051
    
    00:09 16:20
    18:52 26:21
    29:57 34:09
    38:27 43:32
    47:34
    
    16:20                               980579
    YOURS
    343
    00:16:16,209 --> 00:16:17,742       976209 - x 977742 => 4370 before
     They're still alive.
    
    MINE
    344
    00:16:13,239 --> 00:16:14,772       973239 - x 974772 => 974772 + 4370 = 979142 = 16:19
     They're still alive.           
    
    18:52                               1132131
    YOURS
    345
    00:18:57,604 --> 00:18:58,936       x 1137604 - 1138936 => 5473 after
         Do you speak English?      
    
    MINE
    347
    00:18:54,634 --> 00:18:55,967       x 1134634 - 1135967 => 1134634 - 5473 = 1129161 = 18:49
         Do you speak English?      
    
    26:21                               1581680
    YOURS
    521
    00:26:17,442 --> 00:26:19,610       1577442 - x 1579610 => 2070 before
           (knife sharpening)
           
    MINE
    523
    00:26:14,473 --> 00:26:16,640       1574473 - x 1576640 => 1576640 + 2070 = 1578710 = 26:18
           (knife sharpening)       
    
    29:57                               1797095
    YOURS
    543
    00:30:01,032 --> 00:30:03,400       x 1801032 - 1803400 => 3937 after
              (Mayfair)             
        So I need you to send       
         a team to the island       
    
    MINE
    539
    00:29:58,063 --> 00:30:00,430       x 1798063 - 1800430 => 1798063 - 3937 = 1794126 = 29:54
        So I need you to send       
         a team to the island       
    
    34:09                               2049847
    YOURS
    625
    00:34:00,138 --> 00:34:01,371       2040138 - x 2041371 => 8476 before
                  Agh!              
    
    MINE
    621
    00:33:57,169 --> 00:33:58,402       2037169 - x 2038402 => 2038402 + 8476 = 2046878 = 34:06
                  Agh!              
    
    38:27                               2307605
    YOURS
    714
    00:38:42,086 --> 00:38:43,320       x 2322086 - 2323320 => 14481 after
    (man)                           
    Why haWe're not high enough yet.
    
    MINE
    708
    00:38:39,117 --> 00:38:40,350       x 2319117 - 2320350 => 2319117 - 14481 = 2304636 = 38:24
          We're not high enough yet.
    
    43:32                               2612209
    YOURS
    831
    00:43:30,275 --> 00:43:32,342       2610275 - x 2612342 => 133 before
          You're gonna what?!       
    
    MINE
    825
    00:43:27,305 --> 00:43:29,372       2607305 - x 2609372 => 2609372 + 133 = 2609505 = 43:29
          You're gonna what?!       
    
    47:34                               2854051
    YOURS
    868
    00:47:37,521 --> 00:47:40,823       x 2857521 - 2860823 => 3470 after
        (fuselage shuddering,       
         instruments beeping))      
    
    MINE
    869
    00:47:34,552 --> 00:47:37,854       x 2854552 - 2857854 => 2854552 - 3470 = 2851082 = 47:31
        (fuselage shuddering,       
         instruments beeping)       
    
    
    Bottom line of the above analysis is using your data would have worked very well for me:
    MY ORIGINAL
    0: start=00:00 end=16:18
    1: start=18:50 end=26:18
    2: start=29:55 end=34:07
    3: start=38:25 end=43:30
    4: start=47:32

    USING YOUR DATA AND OFFSETS
    0: start=00:00 end=16:19
    1: start=18:49 end=26:18
    2: start=29:54 end=34:06
    3: start=38:24 end=43:29
    4: start=47:31
     
  13. reneg

    reneg Member

    720
    3
    Jun 19, 2002
    That's pretty close, looks approximately a second off. This is certainly cumbersome to analyze by hand. I was comparing with the help of Subtitle Edit & Excel. Subtitle Edit's Synchronization->Point Sync via other subtitle helped a bit (screenshot attached)
     

    Attached Files:

  14. moyekj

    moyekj Well-Known Member

    11,298
    100
    Jan 23, 2006
    Mission...
    OK, now did the reverse. Used my data to calculate what it would end up for you:
    Code:
    0 978144
    1130562 1578977
    1795460 2047178
    2305670 2610541
    2852416
    
    00:00 16:18
    18:50 26:18
    29:55 34:07
    38:25 43:30
    47:32
    
    16:18                               978144
    YOURS
    343
    00:16:16,209 --> 00:16:17,742       976209 - x 977742 => 977742 + 3372 = 981114 = 16:21
     They're still alive.
    
    MINE
    344
    00:16:13,239 --> 00:16:14,772       973239 - x 974772 => 3372 before
     They're still alive.           
    
    18:50                               1130562
    YOURS
    345
    00:18:57,604 --> 00:18:58,936       x 1137604 - 1138936 => 1137604 - 4072 = 1133532 = 18:53
         Do you speak English?      
    
    MINE
    347
    00:18:54,634 --> 00:18:55,967       x 1134634 - 1135967 => 4072 after
         Do you speak English?      
    
    26:18                               1578977
    YOURS
    521
    00:26:17,442 --> 00:26:19,610       1577442 - x 1579610 => 1579610 + 2337 = 1581947 = 26:21
           (knife sharpening)
           
    MINE
    523
    00:26:14,473 --> 00:26:16,640       1574473 - x 1576640 => 2337 before
           (knife sharpening)       
    
    29:55                               1795460
    YOURS
    543
    00:30:01,032 --> 00:30:03,400       x 1801032 - 1803400 => 1801032 - 2603 = 1798429 = 29:58
              (Mayfair)             
        So I need you to send       
         a team to the island       
    
    MINE
    539
    00:29:58,063 --> 00:30:00,430       x 1798063 - 1800430 => 2603 after
        So I need you to send       
         a team to the island       
    
    34:07                               2047178
    YOURS
    625
    00:34:00,138 --> 00:34:01,371       2040138 - x 2041371 => 2041371 + 8776 = 2050147 = 34:10
                  Agh!              
    
    MINE
    621
    00:33:57,169 --> 00:33:58,402       2037169 - x 2038402 => 8776 before
                  Agh!              
    
    38:25                               2305670
    YOURS
    714
    00:38:42,086 --> 00:38:43,320       x 2322086 - 2323320 => 2322086 - 13447 = 2308639 = 38:28
    (man)                           
    Why haWe're not high enough yet.
    
    MINE
    708
    00:38:39,117 --> 00:38:40,350       x 2319117 - 2320350 => 13447 after
          We're not high enough yet.
    
    43:30                               2610541
    YOURS
    831
    00:43:30,275 --> 00:43:32,342       2610275 - x 2612342 => 2612342 + 1169 = 2613511 = 43:33
          You're gonna what?!       
    
    MINE
    825
    00:43:27,305 --> 00:43:29,372       2607305 - x 2609372 => 1169 before
          You're gonna what?!       
    
    47:32                               2852416
    YOURS
    868
    00:47:37,521 --> 00:47:40,823       x 2857521 - 2860823 => 2857521 - 2136 = 2855385 = 47:35
        (fuselage shuddering,       
         instruments beeping))      
    
    MINE
    869
    00:47:34,552 --> 00:47:37,854       x 2854552 - 2857854 => 2136 after
        (fuselage shuddering,       
         instruments beeping)       
    
    YOUR ORIGINAL
    0: start=00:00 end=16:20
    1: start=18:52 end=26:21
    2: start=29:57 end=34:09
    3: start=38:27 end=43:32
    4: start=47:34

    USING MY DATA AND OFFSETS
    0: start=00:00 end=16:21
    1: start=18:53 end=26:21
    2: start=29:58 end=34:10
    3: start=38:28 end=43:33
    4: start=47:35

    So, seems to work pretty darn well to me (for sample size of 1 of course).

    This may be worth pursuing after all. Would be nice to get the data from someone else for this show.
     
  15. innocentfreak

    innocentfreak Active Member

    8,968
    6
    Aug 25, 2001
    Florida
    Downloading now to see what I show.

    This is what my .edl file contains for Blindspot ep 111. Not sure if it is helpful or not.

    0.00 26.39 0
    979.01 1131.00 0
    1580.51 1795.96 0
    2048.58 2306.44 0
    2611.74 2852.88 0
    3361.36 3599.70 0

    This was in the other text document.
    FILE PROCESSING COMPLETE 107883 FRAMES AT 2997
    -------------------
    1 791
    29341 33896
    47368 53825
    61396 69124
    78274 85501
    100740 107883
     
    Last edited: Mar 6, 2016
  16. moyekj

    moyekj Well-Known Member

    11,298
    100
    Jan 23, 2006
    Mission...
    innocentfreak, we would need your .srt file as well.
     
  17. innocentfreak

    innocentfreak Active Member

    8,968
    6
    Aug 25, 2001
    Florida
    I forgot to check captions. I wondered why I couldn't find it.
     

    Attached Files:

  18. moyekj

    moyekj Well-Known Member

    11,298
    100
    Jan 23, 2006
    Mission...
    OK, I have code to automate things now which makes things much easier and more precise than previous postings. Here's my results using each other's data:

    moyekj results
    moyekj CUT DATA
    1: start=00:00:00.000 end=00:16:18.144
    2: start=00:18:50.562 end=00:26:18.977
    3: start=00:29:55.460 end=00:34:07.178
    4: start=00:38:25.670 end=00:43:30.541
    5: start=00:47:32.416

    USING reneg DATA moyekj WOULD GET THIS
    1: start=00:00:07.674 end=00:16:17.609
    2: start=00:18:49.161 end=00:26:18.710
    3: start=00:29:54.125 end=00:34:06.878
    4: start=00:38:24.636 end=00:43:29.239
    5: start=00:47:31.082 end=00:55:59.389

    USING innocentfreak DATA moyekj WOULD GET THIS
    1: start=00:00:25.989 end=00:16:17.508
    2: start=00:18:49.498 end=00:26:19.025
    3: start=00:29:54.475 end=00:34:07.095
    4: start=00:38:24.956 end=00:43:30.239
    5: start=00:47:31.379 end=00:55:59.859

    reneg results
    reneg CUT DATA
    1: start=00:00:09.509 end=00:16:20.579
    2: start=00:18:52.131 end=00:26:21.680
    3: start=00:29:57.095 end=00:34:09.847
    4: start=00:38:27.605 end=00:43:32.209
    5: start=00:47:34.051 end=00:56:02.359
    6: start=00:59:04.040 end=00:59:40.376

    USING moyekj DATA reneg WOULD GET THIS
    1: start=00:00:00.000 end=00:16:21.114
    2: start=00:18:53.532 end=00:26:21.947
    3: start=00:29:58.430 end=00:34:10.147
    4: start=00:38:28.639 end=00:43:33.511
    5: start=00:47:35.385

    USING innocentfreak DATA reneg WOULD GET THIS
    1: start=00:00:27.824 end=00:16:20.478
    2: start=00:18:52.468 end=00:26:21.995
    3: start=00:29:57.428 end=00:34:10.064
    4: start=00:38:27.925 end=00:43:33.209
    5: start=00:47:34.349 end=00:56:02.829

    innocentfreak results
    innocentfreak CUT DATA
    1: start=00:00:26.390 end=00:16:19.010
    2: start=00:18:51.000 end=00:26:20.510
    3: start=00:29:55.960 end=00:34:08.580
    4: start=00:38:26.440 end=00:43:31.740
    5: start=00:47:32.880 end=00:56:01.360
    6: start=00:59:59.700

    USING moyekj DATA innocentfreak WOULD GET THIS
    1: start=00:00:00.000 end=00:16:19.646
    2: start=00:18:52.064 end=00:26:20.462
    3: start=00:29:56.945 end=00:34:08.663
    4: start=00:38:27.154 end=00:43:32.042
    5: start=00:47:33.917

    USING reneg DATA innocentfreak WOULD GET THIS
    1: start=00:00:08.074 end=00:16:19.111
    2: start=00:18:50.663 end=00:26:20.195
    3: start=00:29:55.627 end=00:34:08.363
    4: start=00:38:26.120 end=00:43:30.724
    5: start=00:47:32.582 end=00:56:00.890

    From what I can see from above it seems to work quite remarkably well except using innocentfreak data results in wrong 1st start point which I will need to look into more.
    EDIT: Looking into innocentfreak's 1st start point it appears he intentionally started program at seemingly arbitrary point into the show in his skip data: "You can't have this outside my apartment". So the above results are true to that location so there is no issue.

    NOTE: As part of this system we could easily have a fixed user-defined offset in seconds to apply to all cut points to adjust start points backwards and end points forwards so that you could force partial viewing of commercial segments to avoid show segments being cut off.
     
    Last edited: Mar 6, 2016
  19. innocentfreak

    innocentfreak Active Member

    8,968
    6
    Aug 25, 2001
    Florida
    All I did was download from my Bolt and run metadata, decrypt, qs fix and Ad detect. I forgot to run captions so I ran that separately on the TS file. Weird that it would give different results.

    I can try another episode if you want to see if we get similar results or if it is specific to my recordings.
     
  20. moyekj

    moyekj Well-Known Member

    11,298
    100
    Jan 23, 2006
    Mission...
    OK, so you used comskip to generate the cut file with no manual review. Note that reneg and I manually defined our skip points using VRD so they are expected to be more accurate.
    I don't see anything weird. The differences in results are mostly because of different source cut points to begin with. I think this exercise has proven that syncing cut points by comparing caption files does actually work pretty well, and with the ability to pad either side of cut points we can pretty much guarantee not to miss any actual show assuming the source cut points are accurate to begin with.

    It would be good to pick another show from my list of shows to compare now that I can easily run the comparison using code.

    Setting up an actual online database for sharing these files is probably the next big challenge - I welcome ideas on how to do that and/or volunteers to house the data?
     
    Last edited: Mar 6, 2016

Share This Page